Towards community-scale resilience via repurposable computing continuum infrastructures

1 Introduction

Community-scale Cyber Physical Human Systems (CPHS), an extension of traditional Cyber Physical Systems (CPS) that places the human in the loop as an active participant (Clemmensen et al., 2025), are nowadays increasingly deployed to address diverse societal needs through efforts such as smart cities that efficiently manage resources (water, power, transportation, etc.), deliver services, achieve operational efficiency, and ensure resilience in the face of unexpected events (e.g., disruptive weather and disasters). As with traditional CPS, these CPHS are also typically composed of multiple abstraction levels comprising services, applications, physical infrastructure, and the reactive and dynamic sense-compute-actuate information infrastructure.

Indeed, the interaction with humans and, in many cases, the strong coupling with dynamic environmental conditions make these systems inherently difficult to predict, thereby highlighting dynamic adaptation and system resilience as fundamental design objectives. Despite providing some sort of adaptivity to changing environments, situations and operating conditions, current CPHS are normally purpose-built, designed to operate in their own application context and to react and adapt to changes happening in that context only (e.g., traffic monitoring systems, smart buildings, weather monitoring, etc.). This siloed design paradigm makes the system brittle in the face of extreme events and failures, since individual CPHS are unable to seamlessly interoperate with other/adjacent CPHS to support adaptive responses or joint deployment to tackle emergencies. We posit that repurposing of these CPHS extends the reach of these brittle systems to improve community-scale resilience via integrated, cross-layer adaptations. By repurposing we mean reusing the exact same infrastructure for applications and scenarios that individual CPHS were not originally designed for.

In this work, we adopt the notion of infrastructure resilience as the capacity of community-scale systems to absorb, recover from, and adapt to disruptive events Hosseini et al. (2016) —such as natural disasters, infrastructure failures, or unexpected operational demands—while continuing to deliver essential services to the community. We acknowledge that cross-integration of independently designed systems, while enabling faster responses and resource redundancy, may also introduce new inter-dependencies that could lead to cascading disruptions if not carefully managed. Our position is that the potential benefits of repurposing—namely broader situational awareness, resource sharing, and coordinated response—outweigh these risks when supported by appropriate orchestration, governance, and modularity mechanisms that contain failure propagation across system boundaries.

Such situations manifest in multiple ways: emergency situations; unexpected events; change in the system scope due to the system’s long lifetime; changes in the target applications; changes in regulations; etc. Repurposing of independently designed CPHS facilitates the conjoining of these systems, such that it not only extends overall functionality, but also achieves higher overall system resilience. Furthermore, despite being theoretically possible, repurposing at scale can be achieved only by rethinking the design philosophy of CPHS and by developing, deploying and leveraging several specific enablers. At an high level, policies for achieving CPHS resilience through repurposing can deploy both reactive (event-driven) as well as proactive (anticipatory) mechanisms. Repurposing itself, where these policies are implemented, can be applied at multiple levels of a CPHS (services, applications, and infrastructure) and must ensure the privacy and security requirements of each CPHS. In our opinion, repurposing can be effectively enabled leveraging an intelligent computing fabric–spanning the edge-fog-cloud continuum (Al-Dulaimy and Galletta, 2024). This fabric should be as open as possible to accommodate new applications and made interoperable with other computing infrastructures and devices, and is critically needed to orchestrate efficient and adaptive management of information in a cross-layer manner to meet system objectives, achieve operational efficiency, respond to dynamically changing contexts and situations, and reconfigure to adapt to new applications, goals and constraints.

This article presents a vision for a principled approach for the design of resilient community-scale CPHS, through repurposing existing CPHS via cross-layer self-adaptation.

This vision is informed by over a decade of multiple projects executed by the authors from the EU (Julio de Oliveira Filho, Francesca Palumbo, and Francesco Regazzoni) and the US (Nikil Dutt and Nalini Venkatasubramanian), covering a wide range of large-scale, distributed, CPS and community-scale CPHS systems addressing topics in smart transportation, healthcare, and disaster-based emergency response, with a view to ensuring safe, secure, timely personalized actions in the face of highly dynamic and evolving environments, while also ensuring information privacy across the spectrum of sense-compute-act cycles.

As motivational and running examples, we focus in particular on two scenarios that we have studied and addressed in the past few years.

In the first scenario we present insights gained from the US NSF CareDEX Venkatasubramanian (2022) that informs the design of adaptive systems to deal with extreme events and large disasters by creating actionable information flow under massive failures. This includes information about damage to infrastructure (power lines, bridges, buildings, etc.) and the status of affected citizens, enabling timely and informed responses at the community level, e.g., information assimilation at city governments for targeted alert dissemination and evacuation planning. We notice furthermore, that when the existing deployment can be seamlessly augmented with additional sensing—such as body-worn sensors, drone-based sensing, mobile rovers—to improve coverage of unfolding events and obtain situational awareness, local governments and emergency responders will be better equipped to deliver resources and services where they are most needed. Prior efforts by the team have explored the ability to support perpetual edge-cloud IoT systems for public safety and senior-care applications. (Benson et al., 2015; Uddin et al., 2016; Alhassoun and Venkatasubramanian, 2024).

In the second scenario, we address the efficient and scalable deployment and management of Internet of Thing (IoT) in urban environments to support a wide range of applications. We specifically envision cost-effective hardware and middleware deployments that can enable the repurposing of IoT devices and edge-fog infrastructure used in communities, for example, in road networks and intersections. This instrumentation serves daily operational needs in intelligent transportation systems, traffic management, and human mobility analytics–sensing and analyzing emergent events such as congestion, accidents, etc. Concurrently, it serves also for understanding the impact of weather conditions (flooded or wet road networks, etc.). The data collected from these devices can inform and enhance planning of community services, including vehicle rerouting, traffic scheduling, infrastructure maintenance and upgrades. We argue that Artificial Intelligence (AI)-driven predictive analytics applications in the cloud and at the edge can significantly improve system performance and dependability. During extreme events or natural disasters, such as earthquakes, hurricanes, wildfires, the very same infrastructure is dynamically re-configured to rapidly and reliably meet new monitoring requirements in a failure-prone environment.

A critical requirement for achieving this level of infrastructural flexibility is a robust cloud-to-edge framework capable of supporting multiple applications. The design philosophy behind such a framework is significantly different from the design philosophy currently deployed for designing adaptive CPS. Flexibility can be understood along various abstraction layers:

• Infrastructure Layer: an adequate diversity of sensing/actuation devices, networks, storage and compute resources and hardware accelerators at the edge and in the cloud. But also the capability of adding new resources in a seamless and dynamic way.
• Information Layer: multi-modal information, data management platforms, data analytics including Machine Learning (ML)/AI services, higher-level information nuggets.
• Orchestration Layer: a distributed middleware engine that can manage and repurpose the underlying infrastructure and information resources to effectively execute diverse applications at desired quality levels.

Additionally, the information layer in turn is composed of multiple abstraction layers covering the edge-fog-cloud continuum. The scale, complexity, and dynamicity of these applications creates the urgent need for an adaptive orchestration scheme that supports a geographically distributed computing continuum. Related to this, we report on insights gained from the Horizon Europe Project MYRTUS ¹ (Palumbo et al., 2025). MYRTUS technologies, in a use-case agnostic manner, aims at laying the basis to infer generic heterogeneous computing infrastructures upon which to provide a seamless orchestration of applications in the edge-cloud continuum. The project proposes a generic system architecture that introduces this degree of adaptability–such as the need for a unified containerization and deployment technology suitable for cloud and edge alike, and the use of distributed knowledge bases across layers.

Furthermore, with humans in the loop and the need for adherence to laws and regulations related to management of data, privacy and security issues must be baked into system conceptualization from the beginning. We discuss this angle of self-adaptive systems by summarizing the current related efforts in design and deployment of suitable privacy preserving primitives–including the ones carried out in the Horizon Europe project SECURED ² (Regazzoni et al., 2024) and discuss how security primitives can be deployed reliably overall through the computing continuum, ensuring the required level of protection and providing adaptivity.

The rest of the paper is organized as follows. Section 2 summarizes the two motivating CPHS scenarios. Section 3 outlines the benefits of repurposing, and illustrates examples of repurposing across existing CPHS, Section 4 describes the core technologies required for repurposing in the computing continuum, and Section 5 concludes with the current status and research challenges.

2 Exemplar case studies

In this section we showcase two purpose-built community-scale CPHS exemplars: 1) human-centric sensing for health and wellness in a community setting, together with secure information-exchange mechanisms for healthcare resilience in extreme events such as disasters, and 2) monitoring, analysis and control of community-scale urban mobility through end-to-end orchestration of the edge-fog-cloud continuum. We will discuss the repurpose of the CPHS developed in these use cases in the next sections.

2.1 Case study 1 - CareDEX: rapid and secure information exchange for resilient senior care

In this case study, we describe our prior and ongoing efforts in community healthcare IoT deployments and platforms for healthcare resilience in extreme events such as emergencies and disasters. Healthcare resilience in the face of disasters and emergencies is a global concern. Older adults are highly vulnerable in emergencies-they suffer a disproportionate level of fatalities and injuries during emergencies, extreme events and natural disasters. Despite progress in creating age-friendly communities that promote independent living and improve longevity, continuity of care and safety for older adults during disasters is often limited. Older adults are often unable to shelter safely in place or evacuate on their own due to a range of physical conditions (need for life-sustaining equipment, impaired mobility) and cognitive afflictions (e.g., dementia, Alzheimer’s). Older adults living in retirement communities, senior health facilities (SHFs), and congregant living face increased morbidity and mortality challenges in multi-hazard situations as experienced during the COVID-19 pandemic.

In the CareDEX project (Venkatasubramanian, 2022; Allred et al., 2023; Kenne et al., 2024) (See Figure 1), we developed techniques and mechanisms to create information preparedness to build disaster resilience for our aging communities. Seamless access to information about living facilities (e.g., operational status, number of residents, floor plans) and about residents and their personalized needs (e.g., health conditions such as the need for dialysis, oxygen, wheelchair or assistive devices, and personal objects to reduce anxiety) is typically held by caregivers at the facility; this knowledge can empower first responders to improve response outcomes and safe evacuations during disasters. Such information is sensitive and must be held in ways that protect the privacy of the vulnerable population (older adults) and their caregivers. The acquisition and access to such situational information about the facility (floor plans, operational status) and its individuals (health profiles, care instructions) can guide responders to triage, shelter-in-place or evacuate older adults appropriately, improving eventual outcomes.

Accurate situational awareness is required for timely response - individuals, caregivers, facility leadership and staff, emergency managers, and first responders maintain and access community-contributed information using emerging IoT and big data technologies. Acquisition and timely access to situational information about the facility (floor plans, operational status) and its individuals (health profiles, care instructions) can guide responders to triage, shelter-in-place or evacuate older adults appropriately, improving eventual outcomes. Such information must be gathered apriori, updated regularly, and queried to meet monitoring needs.

The CareDEX Platform includes a suite of decision support tools that addresses specific needs associated with the rapid exchange of personalized care information between first responders, caregivers in senior housing facilities, and older adults. The platform encompasses mechanisms to gather data about in-facility components which ensures caregivers have access to personalized information about residents’ care needs, health conditions and facility resources and regional components which provides situational awareness during imminent disasters to ensure smooth transfer for older adults to safety.

The in-facility component of CAREDEX leverages the smartspace IoT data management middleware, TIPPERS Mehrotra et al. (2016), Lin et al. (2021) to create the smart spaces and applications required for CareDEX. In that sense, we view this CAREDEX as a smartspace extender technology atop the TIPPERS platform Mehrotra et al. (2016). CareDEX leverages the availability of smart-space technologies to help with the creation of a digital twin that captures and exchanges relevant spatial, temporal and human aspects in a virtual representation of the physical environment. The platform also empowers the creation of novel applications, such as localization, space management, and resource allocation, which enhance the capabilities of the smart space platform. The Data Representation Module leveraged the TIPPERS environment to create and manage representation of spatial data and resident information and populate space, resident and sensor information table to the database. The Ingestor Module in CareDEX enabled dynamic ingest of heterogeneous data from many data sources, including live sensor readings. and inserts this into the appropriate database structures for querying and analysis. Custom integration mechanisms were designed for new devices that were incorporated into the drills (smart watches, WiFi tags, etc.) and tested. The Data cleaning and analytics component was used to extracting spatial and resident information to generate accurate information about the occupancy of spaces and their use, localization of individuals and their health, etc. The CAREDEX Dashboard provided Data visualization through key applications developed through pre-established queries. By promoting effective coordination and support, CareDEX empowers first responders and caregivers to improve response outcomes, prioritize resources, and provide tailored care to older adults, ultimately enhancing their resilience in crisis situations.

The Regional Impact Module is an independent but connected CareDEX tool allows us to create situational awareness at the regional level. The regional impact module is an interactive online visualization tool that allows users to view the impact of natural disasters at regional scale. The goal of the tool is to provide information to non-technical user about 1) where has the disaster event occurred and how close is it? 2) what type of damage could be expected in areas nearby? The application frontend is desktop SPA developed using vuejs as the core framework, with Leaflet and GoogleMaps libraries providing map functionalities. The backend API was developed with the Django REST framework with data stored in a PostgreSQL database with GIS support. As part of the backend, there is a component that regularly monitors several public sources of hazard data and automatically downloads the necessary information. For each exercise, appropriate datasets were incorporated and useful dashboards were generated in coordination with the civic partners. Over time, the regional tool was further expanded to integrate more specific information about the older adults and used for planning and conducting the evacuation. Examples include information about earthquakes based on the HAZUS toolkit(USGS), wildfire information from the National Interagency Fire Center, NOAA’s Hazard Mapping System Fire and Smoke Product ³ . AIRNOW ⁴ has hourly data update of measures of ozone, PM1.0 and PM2.5 in the air.

2.2 Case study 2 - urban traffic monitoring infrastructure in communities

Urban Traffic Monitoring (UTM) systems (Chowdhury and Sadek, 2003; Vlahogianni et al., 2014) typically combine a heterogeneous set of sensors—including cameras, inductive vehicle detection loops, microphones, and crosswalk buttons—with traffic actuators such as traffic lights, Accessible Pedestrian Signals (APS), and traffic barriers. These components are distributed over large, densely populated urban areas. The sensors continuously generate data streams that feed an increasing number of software services deployed either locally at the edge or in the cloud. The primary objectives of these services are to monitor traffic conditions and regulate traffic flow in urban environments.

The EU project MYRTUS (Palumbo et al., 2024) aims to demonstrate the benefits of end-to-end orchestration across the cloud-to-edge continuum when applied to the monitoring, analysis, and control of mobility systems. The project is deployed at a traffic-intensive urban intersection in the city of Helmond and covers two adjacent intersections. Traffic lights at this intersection are equipped with smart cameras, edge computing devices, SW services, and communication equipment (see Figure 2). This infrastructure enables the detection and tracking of vehicles and other vulnerable road users, such as groups of pedestrians or cyclists. Based on these observations, the system can estimate collision risks and issue warnings to approaching vehicles.

All sensors and actuators are connected to an embedded Jetson Orin device (NVIDIA Corporation, 2025), which is physically installed in the roadside traffic control cabinet and is intended to operate as a fog server. Depending on operational requirements, selected processing tasks can be enabled or disabled on this device, such as vehicle detection from camera images. All components are interconnected through a dedicated subnetwork, and secure VPN connections provide access to cloud resources. Authorized services can interact with the deployed devices and consume their data, as illustrated in Figure 3.

Within the MYRTUS project, particular attention is given to the dynamic orchestration of containerized software functions for road user detection and trajectory monitoring across the edge, fog, and cloud layers. As shown in Figure 4, one container exposes Application Programming Interfaces (APIs) to access embedded Deep Learning (DL)accelerators and camera image streams, while a second container hosts the road user monitoring application itself.

Within MYRTUS, this use case considers three operational modes. In the first mode, road user detection is performed entirely within the cameras, and the resulting messages are transmitted directly to test vehicles approaching the intersection. In the second mode, part of the processing is delegated to the fog server, including the fusion of detections from multiple cameras and, when required, road user detection based on blurred video streams to preserve user privacy. In the third mode, blurred video streams are forwarded to the cloud, where full processing can be executed, for instance to derive long-term statistics and characterize traffic usage patterns over extended time periods. An example of an application that tracks different traffic participants—pedestrians, cyclists, and motor vehicles—is shown in Figure 5.

MYRTUS works on a framework for orchestration and deployment in the edge-cloud continuum. The research in this use case allows the whole set of software resources to be move between edge and cloud computing resources, scale the usage of available sensors, and is an open platform for future expansion. As we discuss in detail further in Section 4, we see this piece of technology as an enabler for community-scale repurposing of such infrastructures.

3 Repurposing in the cyber-physical-human continuum

Existing edge-cloud IoT deployments are typically engineered for specific, narrowly defined goals, with hardware/software stacks tightly integrated to well-specified operational constraints. Repurposing an existing infrastructure for applications it was not originally designed to support introduces substantial complexity. In such scenarios, computing resources, networks, sensors, and actuators must be reused and adapted in an opportunistic manner (combining reactive and proactive mechanisms), exploiting the installed base to the greatest extent possible while minimizing physical modifications, manual intervention, and custom reconfiguration. This approach seeks to maximize the utility of existing infrastructures while maintaining acceptable performance and reliability for both the original and the newly introduced applications.

In this section, we first articulate the benefits of repurposing across the cyber-physical-human continuum. We then illustrate some of these benefits through the CareDEX and MYRTUS exemplars presented in Section 2, showing how large-scale ecosystem disruptions (e.g., natural disasters, coordinated attacks, or anomalous events) impact independently deployed, siloed platforms and motivate cross-system adaptation.

3.1 Benefits of repurposing

Infrastructure and application repurposing can result in an integrated, coupled platform that can more effectively adapt to changing environment and situations. Such repurposing can occur both across layers and across systems, with the coupled system providing a number of technical, economic, and societal benefits across the computing continuum:

• Rapid adaptation to emerging requirements: Repurposeable infrastructures can be quickly adapted to new applications, workloads, or operational goals—such as emergency response or temporary community services—without lengthy redesign or redeployment cycles.
• Increased resilience and fault tolerance: The ability to reassign workloads and reconfigure service chains across layers enhances system resilience, allowing infrastructures to continue operating despite failures, congestion, or partial resource unavailability.
• Better support for safety-critical and community-oriented applications: In scenarios such as disaster response, healthcare, or public safety, repurposing enables rapid activation of decision-support services by leveraging already deployed sensing and computing assets.
• Improved resource utilization: Repurposing enables existing computing, sensing, and networking resources to be reused for multiple applications over time, reducing idle capacity and improving overall infrastructure efficiency across layers.
• Cost reduction and extended infrastructure lifetime: By minimizing the need for new deployments and hardware upgrades, repurposing lowers capital and operational expenditures while extending the useful lifetime of installed infrastructures.
• Support for innovation and experimentation: Repurposing lowers the barrier to deploying and testing novel applications on real infrastructures, facilitating experimentation, prototyping, and technology transfer without disrupting primary services.
• Improved sustainability and environmental Impact: Reusing existing infrastructures reduces hardware duplication and energy consumption associated with new deployments, contributing to more sustainable and environmentally responsible computing systems.

We illustrate the role of joint/coupled repurposing and adaptation using the earlier examples of healthcare and intelligent transportation in the case of an extreme environment (e.g., earthquake, flood) that affects both deployments: (a) Impact on health–evacuation of older adults and those with severe health issues in emergencies and (b) Impact on road networks–Similarly, human behavior (congestion caused by evacuating populations) can require reconfiguration of traffic infrastructure to help ingress/egress of emergency vehicles required to respond to large disasters. Multi-modal sensing/actuation and adaptation can help detect opportunities and avoid pitfalls of the joint deployments.

3.2 Repurposing CareDEX

Through the use of a smart-space platform, CAREDEX enables the assimilation and exchange of customized care information of seniors (health state, medications, assistive devices) rapidly between first responders, caregivers in senior housing facilities, and older adults in a secure manner. CareDEX empowers organizations to readily assimilate, ingest, store and exchange information, both apriori and in real-time, with response agencies to protect and care for the elderly in extreme events. Using CareDEX, caregivers and residents can seamlessly share information with first responders regarding changing health conditions, personalized needs and identify those in need of specialized triage and critical care. Given the sensitive nature of personal information (e.g., health profiles, ability status, medications) and the cybersecurity implications of data sharing, CareDEX incorporates techniques to balance the need for individual privacy with authorized release of information to responders when needed. CareDEX addresses user data privacy and cybersecurity by virtue of its underlying middleware platform.

While CareDEX was originally designed as a rapid and secure information platform to facilitate safe evacuation of seniors during disasters and emergencies, CareDEX could be repurposed for several community-scale resilience scenarios: coordination of multiple site evacuations (schools, hospitals, offices, concert halls, stadiums, etc.) during a large-scale disaster. Each location has populations with specific needs that must be matched with evacuation destination sites meeting those needs. The status of road infrastructure from MYRTUS must be conjoined with the repurposed CareDEX secure information exchange system to enable safe and efficient city-wide evacuation plans, as described in the next section.

3.3 Repurposing of UTM systems

Under normal operating conditions, UTM systems support traffic flow monitoring, intersection safety, and crowd or congestion management. However, the system’s geographic distribution and its heterogeneous sensing modalities provide a unique opportunity to temporarily adapt or repurpose this infrastructure during extraordinary circumstances or in support of specific community-oriented objectives. During disasters or extreme environment events, this same UTM infrastructure can be repurposed to detect traffic signal and sensor failures, road accessibility and conditions, and their impact on transportation corridors–such as traffic pileups and major accidents. In combination with the CareDEX infrastructure, this information can be valuable in the planning of evacuation, area isolation, or transport of wound or elderly people.

Damage to the infrastructure, unexpected congestion in road networks, or delays can significantly degrade road navigability and may cascade into widespread regional traffic disruptions. The infrastructure deployed to monitor urban traffic provides a compelling example of the potential to repurpose CPS for enhanced resilience and situational awareness.

Two representative repurposing scenarios are considered in the following discussion: large-area congestion analysis and multi-camera tracking for critical transportation or crowd management.

In the first repurposing scenario, data generated by traffic cameras, inductive presence loops, and speed measurement devices are aggregated over extended temporal horizons to extract indicators of congestion, such as traffic throughput, average queue length, and vehicular density. These metrics can be correlated across multiple intersections and further enriched with contextual information derived from municipal calendars, including planned events, public works, or seasonal variations. Such integrated analysis supports the characterization and prediction of congestion phenomena at a city-wide scale. For instance, it enables the assessment of the traffic impact of temporary road closures or openings due to events or maintenance activities, the evaluation of the effectiveness of different congestion mitigation strategies once congestion has formed, and the short-term prediction of emerging congestion hot spots to facilitate proactive rerouting or control measures.

There are some characteristics of this use case which are worth noting. In this case, the secondary application can reuse and augment data already collected by the primary traffic-monitoring function. In this sense, the infrastructure is not entirely diverted from its original purpose; rather, its functionality is extended through the combination and higher-level exploitation of existing data streams. This adaptation implies a partial co-existence of regular, infrastructure targeted applications and secondary goals. Aspects such as performance, security, safety must be prioritized and guaranteed first for the primary goal, and shared or coordinated when possible for other uses. These aspects motivates flexible, adaptive mechanisms for orchestration, for example,. We discuss these requirements further in Section 4.1.

The second repurposing scenario illustrates a more substantial retargetting of the UTM infrastructure, in which the primary traffic-monitoring function is temporarily suspended. The multiple cameras deployed at each intersection can be coordinated to form a large-area visual tracking system capable of automatic target recognition and seamless handover of tracked entities from one camera to another. Such a system is particularly relevant for the monitoring of critical transports, including medical convoys, high-profile individuals, or other sensitive resources that require enhanced situational awareness as they traverse the urban environment. It may also support police or emergency operations in which traffic flows must be tightly controlled and dynamically adjusted. In addition, the same capabilities can be leveraged for crowd monitoring during large-scale public events, such as parades, demonstrations, or festivals, where the accumulation and movement of people need to be observed to ensure public safety.

In these scenarios, the UTM infrastructure is employed in a manner that diverges significantly from its original design objectives, demonstrating its potential flexibility and value as a shared urban sensing resource under exceptional conditions. When these conditions are in place, a large extent of the initial requirements on performance, security, and safety change. For example, access rights to networks, data, and computing resources can shift. This motivates flexible, adaptive mechanisms for governance. We discuss these requirements further in Section 4.1.

4 Technologies for repurposing in the computing continuum

The underpinning concept enabling facilitated large-scale system repurposing is the computing continuum, which will be discussed in this section. In fact, effective repurposing in large-scale systems could be achieved by leveraging hierarchical and distributed computing infrastructures, where.

• Computation, storage, and networking resources are flexibly distributed across cloud, fog, edge, and often IoT endpoints; and
• Applications are reconfigured through an appropriate middleware and dynamically moved in the continuum space.

From an infrastructure perspective, cloud computing remains the foundation for large-scale storage and high-performance processing. Platforms such as Microsoft Azure IoT Hub (Microsoft, 2026), together with solutions from Amazon (Amazon Web Services, 2026) and Google (Google Cloud, 2026), provide end-to-end IoT connectivity and cloud-based analytics. In Europe, initiatives such as Gaia-X (Baur, 2025; Gaia-X European Association for Data and Cloud, 2026; Geilenberg et al., 2024) promote federated and secure cloud infrastructures aligned with EU data protection and data sovereignty requirements.

At the other extreme of the computing landscape, Eclipse Kura (Eclipse Foundation, 2026) is an example of an open-source IoT edge framework (in this case, considering the IoT edge as sensing nodes with processing capabilities) designed to simplify the development, deployment, and management of IoT applications. Fog computing, originally introduced by Cisco as an intermediate layer between cloud and edge, brings computation and storage closer to data sources to reduce latency and communication overhead (Bonomi et al., 2012). The OpenFog Reference Architecture, jointly defined by the OpenFog Consortium and the IEEE Standards Association, was later standardized as IEEE 1934 (IEEE Standards Association, 2025). At this level, FogBus2 (Deng et al., 2021) represents a lightweight, blockchain-based framework integrating fog, edge, and cloud resources, offering data integrity, platform independence, and energy efficiency, although it still remains currently a research solution.

Building on initially siloed solutions, several state-of-the-art efforts have attempted to integrate computing layers across the cloud–edge spectrum. CEFIoT (Javed et al., 2018) represents an early cloud–edge framework that combines Raspberry Pi–based edge clusters with public cloud services, leveraging Kubernetes for fault-tolerant management and dynamic reconfiguration. However, it primarily emphasizes fault tolerance and does not address heterogeneity or workload adaptability. Intel’s IoT Reference Architecture (Intel, 2023) integrates edge and cloud resources through an intelligent network layer and has achieved strong industrial uptake, but its reliance on Intel-specific technologies and cloud ecosystems limits flexibility and raises vendor lock-in concerns. OpenStack (OpenStack, 2026), while not explicitly framed as a computing continuum solution, supports unified cloud–edge infrastructures with mechanisms for communication, state synchronization, and lifecycle management over unreliable links. Collectively, these approaches advance the notion of a compute continuum, yet remain largely homogeneous and limited in compositional flexibility.

Indeed, a generalized solution for computing continuum infrastructures is not there yet, while it is true that different communities such as HiPEAC (High Performance, Edge And Cloud computing) and AIOTI (Alliance for Internet of Things Innovation) are actively contributing to formalize terms and concepts behind it as well as to define the open challenges, such as defining and managing infrastructures capable of effectively supporting repurposing. It is worth mentioning that two different initiatives, the EU-CEI (European Cloud, Edge and IoT Continuum Initiative) (EU Cloud Edge & IoT Initiative, 2026) and DISCOVER-US (HiPEAC, 2026), are doing concrete steps to identify and tackle those open challenges, investing respectively in standardization promotion (EU Cloud Edge & IoT Initiative, 2026) and in international combined actions among US and EU Vermesan, (2025).

Considering the management perspective, the lack of integrated heterogeneous continuum infrastructures reflects on the lack of cross-layer management strategies, limiting the optimization potentials and the possibility of effective runtime adaptivity to changing conditions, and thus of repurposing capabilities. Kubernetes Kubernetes, (2026) is currently the open-source de facto standard to automate the deployment, scaling, and management of containerized applications for modern cloud-native applications. Based on Kubernetes, ArubaKube distributed Liqo (Arubakube Cloud, 2026) enables seamless, peer-to-peer multi-cluster management, allowing workloads to be offloaded and managed across different cloud regions. Recently, in the context of the MYRTUS project, this tool has been stretched to handle multi-layer infrastructures. The use of containers and container management Hong et al. (2017); Li et al. (2024) in edge-cloud community settings for energy-aware and mission-critical infrastructure deployments have been explored in recent smart and connected community projects (Venkateswaran et al., 2021; Bouloukakis et al., 2021).

Despite the availability of advanced approaches for edge–cloud orchestration (Hasan et al., 2024; Chen et al., 2024; Tong et al., 2020), edge-cloud continuum models (Pereira et al., 2022) and adaptive Kubernetes-based workload scheduling techniques (Robles-Enciso and Skarmeta, 2024), resources across computing layers are still largely managed as isolated silos. In practice, these solutions have not yet reached the level of maturity required for large-scale deployment, primarily due to the inherent heterogeneity of computing environments and resource types across the continuum.

4.1 Key enablers for repurposing and adaptation

In this section, we discuss a non-exhaustive set of mechanisms that are either required or highly beneficial for the design of computing continuum infrastructures with sufficient flexibility to support systematic application and infrastructure repurposing. The objective of these mechanisms is to enable an installed system to be dynamically adapted to new applications, operational contexts, or stakeholder requirements, without extensive redesign or physical redeployment. Such adaptability is essential for infrastructures spanning the edge–cloud continuum, where hardware capabilities, software stacks, and operational constraints differ significantly across layers.

Several of the mechanisms discussed here are informed by challenges and practical experience gained in recent research and innovation projects, including MYRTUS and CAREDEX, which address heterogeneous, distributed, and safety-critical CPS. These projects highlight that achieving infrastructure repurposability imposes fundamental requirements on both the software and hardware layers. These requirements stem from the need to support heterogeneous execution environments, evolving application chains, and dynamic operational constraints, while still providing acceptable guarantees in terms of performance, security, reliability, and regulatory compliance.

Based on the authors’ experiences, the following requirements can be identified as particularly relevant:

• Containerization Lightweight and flexible containerization technologies, explicitly targeting edge and embedded nodes with limited resources and heterogeneous computing nodes.
• Seamless deployment Deployment mechanisms that operate consistently across edge, fog, and cloud layers, and across heterogeneous Operating Systems (OSs) and middleware stacks.
• Intelligent orchestration Orchestration mechanisms that automate discovery and placement of software components across layers–complying with resource, security, and communication constraints.
• Security and trust Security primitives and governance mechanisms that remain effective when infrastructures are repurposed, federated, or shared across organizational boundaries.

Across the computing continuum, these requirements manifest differently at the edge, fog, and cloud layers. Cloud platforms generally provide more mature abstractions, standardized interfaces, and automation capabilities. In contrast, edge and fog environments remain constrained by limited resources, heterogeneous hardware, and fragmented software support. Bridging these gaps requires coordinated solutions that operate both within individual layers and across layers. In the following subsections, we discuss these requirement and enabling technologies in greater detail, highlighting existing support as well as remaining gaps at each layer of the computing continuum.

4.1.1 Containerization

Container-based virtualization has emerged as a dominant modularization and deployment strategy in cloud computing systems. Technologies such as Docker (Rodriguez and Buyya, 2019; Docker, 2026), Kubernetes (Senjab et al., 2023; Kubernetes, 2026), and Podman (Husain et al., 2023; Podman, 2026) have significantly simplified the deployment and management of complex software systems by addressing challenges related to scalability, dependency management, integration, and inter-component communication. Containers encapsulate applications together with their runtime dependencies, thereby mitigating environment-specific issues and enabling software redistribution and migration—both essential for application repurposing.

Despite their widespread success in data-center and cloud environments, existing container technologies often make implicit assumptions about the underlying computing infrastructure, including the availability of standard processor architectures, ample persistent storage to host relatively large container images, and stable, high-bandwidth connectivity that supports communication mechanisms such as REST-based APIs over Unix sockets. These assumptions break down at the edge and fog layers, where nodes are often resource-constrained, highly specialized, intermittently connected, and based on non-mainstream architectures. As a result, while fog micro-data centers—which largely resemble small local clouds—benefit from established container-based solutions (Pérez de Prado et al., 2020), integrating such solutions on edge nodes often requires significant manual configuration and customization, particularly when managing heterogeneous accelerators such as Graphics Processing Units (GPUs) or Field-Programmable Gate Arrays (FPGAs).

To address these limitations, container technologies must be extended to support a broader range of processor architectures and heterogeneous computing nodes, constrained storage footprints, diverse OSs, and non-traditional communication mechanisms. In particular, container solutions explicitly designed for edge environments are a key enabler for repurposable CPS and CPHS, as they allow already deployed devices to be dynamically reconfigured and reassigned to new tasks. This capability is fundamental in community-scale scenarios, where installed sensing and computing resources may need to be temporarily reassigned to new societal functions.

In this context, a fundamental research direction concerns the extension of containerization to accelerator-based platforms, especially heterogeneous edge devices equipped with reconfigurable hardware such as FPGAs. These platforms target the acceleration of computation-intensive workloads, increasingly driven by AI models, while improving energy efficiency (leveraging hardware specialization) and system flexibility (leveraging different forms of reconfiguration) (Kalapothas et al., 2022). In this context, container orchestration can be managed by leveraging a lightweight OS running on the CPU of heterogeneous nodes. MYRTUS will shortly deliver a custom minimal Linux distribution that integrates Liqo together with the required drivers and software components for containerized workload management on heterogeneous edge nodes.

4.1.2 Seamless deployment

The adoption of containerization has enabled major advances in the deployment of distributed application chains within homogeneous computing clusters. Orchestration frameworks such as Kubernetes provide standardized mechanisms for deploying, scaling, and managing containerized workloads on pools of machines with relatively homogeneous characteristics, typically sharing similar hardware architectures and OSs. These frameworks provide abstractions for service discovery, load balancing, and fault tolerance, thereby reducing the operational complexity of large-scale deployments.

Extending these deployment paradigms across the full computing continuum remains an open challenge. Edge–fog–cloud infrastructures exhibit deep heterogeneity in computational capabilities, OSs, network connectivity, and middleware. Seamless deployment in this context requires mechanisms that operate across layers, manage cross-layer dependencies, and adapt deployment strategies to the capabilities and constraints of each node. Overcoming these challenges is a prerequisite for enabling truly flexible and repurposeable infrastructures, as it allows application components to be be dynamically relocated or reconfigured in response to new goals and purpose.

The MYRTUS project addresses these challenges through the development of a generic architectural concept aligned with emerging EU-CEI standardization initiatives (Palumbo et al., 2025). Key elements under research are:

• A Resource Registry component and associated data models for publishing, reporting, and discovering computing resources. The registry provides a unified cross-layer view of infrastructure capabilities, enabling the systematic (re-)matching of software service requirements with available resources. This functionality is particularly important when re-targeting existing infrastructures or deploying secondary service chains, as it offers a consistent and authoritative source of information about system configuration.
• A deployment mechanism spanning across continuum, designed to preserve service communication semantics while enforcing governance and security constraints during deployment adaptation. Such capabilities, leveraging Liqo virtualization capabilities, are intended to reduce the effort required to transfer software components to platforms with different architectural characteristics, thereby facilitating infrastructure repurposing.
• Distributed resource monitoring and workload characterization mechanisms that complement the static resource descriptions, included within the Resource Registry, with dynamic information about actual resource utilization and workload behavior. While not strictly required for basic deployment, in particular when dynamic workload execution optimization or repurposing are not to be supported, these mechanisms are critical inputs for intelligent orchestration, as discussed in Section 4.1.3.

Cloud layers today provide mature Continuous Integration and Continuous Deployment (CI/CD) pipelines built around widely adopted standards, particularly Kubernetes (Saleh et al., 2025). However, as deployment shifts toward the edge, growing heterogeneity in hardware, OSs, and runtime environments results in fragmented tooling and reduced automation, even if custom solutions start to appear (Ulzurrun et al., 2024). Fog environments offer partial orchestration and CI/CD support (Awad et al., 2025), while edge layers often rely on manual configuration and platform-specific solutions. Bridging this gap is critical to enabling systematic and scalable application repurposing across the computing continuum.

4.1.3 Intelligent orchestration

Effectively mapping software components to available resources—particularly in large-scale and heterogeneous architectures—requires not only automated, but also informed, orchestration mechanisms that are aware of cross-layer constraints. Such mechanisms, as discussed in Section 4, are still not available at the state of the art. The MYRTUS project addresses this need through the development of a cognitive orchestration engine. This engine builds upon information provided by the resource registry, distributed resource monitoring, and workload characterization mechanisms discussed in Section 4.1.2, enabling the optimized placement of service chains across distributed, multi-layer computing systems. The orchestration process enforces strict matching between service requirements and resource capabilities, while also accounting for non-functional constraints such as privacy policies, security requirements, and geo-location restrictions. In many cases, multiple feasible deployment configurations satisfy these constraints; MYRTUS therefore investigates the use of ML techniques and distributed optimization methods to efficiently explore the resulting design space and to identify solutions that are near-optimal with respect to performance, energy efficiency, or resource balancing.

In MYRTUS, cognitive orchestration primarily optimizes workload execution. However, by implementing an autonomic self-adaptation loop, it also becomes a key enabler for adapting and repurposing infrastructure toward new application domains, such as community-scale services activated during emergencies or other exceptional events. Without intelligent orchestration, such adaptations would depend on ad hoc, manual solutions that scale poorly and lack robustness. In contrast, automated, knowledge-driven orchestration potentially reduce development effort and deployment time, while lowering costs by abstracting much of the inherent complexity of large, heterogeneous infrastructures.

Intelligent orchestration also creates the need for a new generation of urban planning tools with the ability to seamlessly integrate new devices and software. For instance, SmartParcels Chang et al. (2025) is an early example of a planning tool that augments communities with newly added IoT and software. This planning approach distinguishes between the smartspace infrastructure layer (sensing, networking, and computing devices) and the associated information units (data and analytics) to support a range of applications by treating applications, information/data, infrastructure, and geophysical layout as interdependent layers. Planners and decision makers can explore design options in a plug-and-play manner to explore trade-offs like cost optimality and sensing coverage. SmartParcels offers multiple design choices: clean-slate that designs communities from scratch, and retrofit to support reusability of existing infrastructure in design. Such planning tools would become critical for repurposing infratructure and information flows across applications.

4.1.4 Security and trust

Given the scenario where multiple, separately managed, intelligent infrastructures must be repurposed by deploying existing platforms and exchange information towards a common mission, several critical issues in security and trust need to be addressed. We motivate these through the previously described large-scale disaster scenario that has caused significant damage to a community and its infrastructures. In this context, healthcare facilities such as senior homes, rehabilitation centers, and hospitals (as in the CareDEX scenario described earlier) must now relocate residents to other safe destinations where they can be triaged and cared for. Secure and protected access to in-facility and patient information is therefore required–patients may need to be transported to varying destinations based on their current and predicted medical conditions. However, these data are typically protected and cannot be shared outside the originally granted circle of trusted users of the data (e.g., caregivers, care facility personnel). New privacy and security challenges arise from the need to temporarily grant access to first responders and disaster planners for selected subsets of this confidential data (e.g., medical condition, special needs). The duration of this temporary access must be only as long as is needed to complete successful evacuation to safe destinations, after which the access to this data shoud be revoked. Furthermore, no private data should persist after the evacuation processs. This requires dynamic access control policies, and mechanisms to facilitate data provenance and traceability to ensure trust in the overall system. Similarly, Situational Awareness (SA) is required to assess the impact of the disaster at the regional scale. Such region-wide SA requires detailed information about which alternate facilities are operational, the availability of space and critical resources (e.g., medical supplies, rehabilitation devices, power, water, etc.) at these destinations, as well as the condition of road networks to help transport patients to the selected destinations. Similar, but different privacy and security needs arise when exchanging confidential information between organizations for creating actionable SA, which in turn can help save lives and property in extreme conditions.

Sharing data between organizations is critical for better SA which in turn can help save lives and property in extreme conditions.

5 Current status and research challenges for repurposing the next-generation CPHS continuum

Figure 6 visualizes the integration middleware needed to support cross-infrastructure repurposing, that ensures privacy-aware and secure data exchange. In the following, we first summarize the current state of repurposing enablers and then discuss some of the ongoing and open research challenges for facilitating repurposing of the next generation CPHS computing continuum.

5.1 Status of repurposing enablers

Table 1 summarizes the level of support for the key repurposing enablers discussed in the previous section across the edge, fog, and cloud layers of the computing continuum. As anticipated, while cloud platforms already provide mature support for most mechanisms—often at the expense of specialization and energy efficiency—significant gaps remain at the edge and fog layers. These layers are precisely where repurposing is often most valuable, due to their proximity to the physical environment and the presence of extensive installed infrastructures. What is certain is that, while individual mechanisms such as containerization, deployment automation, and orchestration are essential enablers for application repurposing, they are not sufficient in isolation. Interoperability across layers and devices—enabling applications and services to move seamlessly between edge, fog, and cloud environments without being tightly coupled to specific vendors, platforms, or middleware stacks—remains largely unrealized and continues to represent a major open challenge. A major obstacle to such interoperability is the prevalence of vendor lock-in, which manifests at multiple levels: e,g. hardware accelerators with proprietary toolchains, platform-specific APIs for deployment and orchestration, and closed monitoring and telemetry solutions. All these constraints significantly limit the ability to repurpose infrastructures, as software components and deployed service chains become bound to specific execution environments and cannot be easily moved elsewhere, in case of infrastructure discontinuity.

TABLE 1. Support for application repurposing enablers across the computing continuum.

Additionally, from a repurposing perspective, beyond the two classical dimensions of interoperability—namely horizontal interoperability, which enables applications to be deployed across heterogeneous devices within the same layer, and vertical interoperability, which allows application components and data flows to migrate or be restructured across edge, fog, and cloud layers—a third dimension becomes essential. Organizational interoperability supports federation across infrastructures owned and operated by different stakeholders and is fundamental in scenarios such as disaster response and community-scale services. Indeed, as we have seen in the previous section, current technologies partially address these challenges. Container standards (Docker, 2026), declarative deployment models (Kubernetes, 2026), and infrastructure-as-code (Rahman et al., 2019) approaches provide a baseline for portability, primarily in cloud and fog environments. Service mesh technologies and standardized communication protocols further decouple application logic from deployment location. However, in highly heterogeneous environments and especially at the edge, interoperability is still limited by device-specific drivers, proprietary accelerator stacks, and inconsistent management interfaces.

To overcome vendor lock-in and enable effective cross-layer repurposing, several research and engineering activities are ongoing, as listed below.

• Extensible resource description models are required to abstract device capabilities—such as compute, accelerators, networking, and sensing—in a vendor-neutral manner. Recently, the members of OASIS Open (OASIS Open, 2026a), the global open source and standards organization, have approved the Topology and Orchestration Specification for Cloud Applications (TOSCA) Version 2.0 (OASIS Open, 2026b) as an OASIS Standard. The goal is to bring TOSCA to become a universal orchestration language applicable across virtually any domain, from traditional IT infrastructure to IoT deployments, edge computing, and industry-specific automation requirements. Nonetheless, it will take time before an effective adoption at the far edge becomes a reality.
• standardized APIs for deployment, monitoring, and orchestration must be adapted to resource-constrained environments without sacrificing expressiveness.
• Policy-aware mediation layers are needed to reconcile differences in security, privacy, and governance requirements when services span multiple layers and administrative domains. An example of such policy mediation layer are the Trust Models Gaia-X, (2025) of GAIA-X. GAIA-X Trust Models provide a structured framework for establishing verifiable trust relationships across federated infrastructures. By defining common specifications for identity management, credential verification, data sovereignty, and compliance assurance, GAIA-X enables interoperable yet policy-constrained resource sharing among independent stakeholders. Incorporating such trust mechanisms into cross-layer mediation layers would strengthen transparency, enforce governance constraints programmatically, and support secure repurposing of services across organizational boundaries.

MYRTUS has already taken concrete steps to tame interoperability by separating infrastructure description, deployment logic, and orchestration intelligence, thereby reducing platform coupling. Nonetheless, as summarized in the last row of Table 1, repurposeability readiness is yet to be achieved.

5.2 Role of AI mechanisms

The integration of AI methods to extract live situational awareness and long-term impacts is a promising direction given the rapid growth in edge AI methods.

Federated Learning (FL): FL appears to be a promising approach to create robust models, where participants (individuals or care facilities) process the raw data onsite, and share models that can be fused at a centralized aggregator. The FL approach supports privacy needs by allowing AI/ML techniques to process sensitive (e.g., healthcare) data locally without sharing the data itself. FL aims to train ML models using data from multiple devices. The heterogeneity of health data from various devices presents a significant challenge in developing robust healthcare models. In the medical field, rare diseases or atypical health states are often underrepresented, while normal or common health indicators are overrepresented. This imbalance can lead to skewed models that may not accurately detect or respond to critical but infrequent health events. For instance, in heart rhythm monitoring, normal sinus rhythms might dominate, while potentially life-threatening arrhythmia’s appear infrequently. Similarly, in fall detection systems for elderly care, the vast majority of data might represent normal movement patterns, with actual fall events being rare occurrences. This data imbalance can result in models that excel at recognizing common patterns but fail to identify crucial anomalies or health emergencies. An early approach to this is exemplified by FLIPS Bhope et al. (2023), a middleware system to manage data and participant heterogeneity in FL training workloads. FLIPS addresses this challenge by clustering devices/participants based on their data distributions. This way, similar devices/participants are grouped together and devices/participants across different clusters are chosen equitably to improve the diversity of the trained model, which improves accuracy and reduces communication costs. FLIPS also uses a trusted execution environment to protect the privacy of participants sharing their data distributions. In the context of CAREDEX, FLIPS can train heart rhythm and fall detection models without compromising sensitive personal data, enhancing both privacy and model robustness.

Handling Scenario Shifts with AI: Traditional FL assumes a distribution of data with a static model. However the scenarios we envisage result in both the data and the models to evolve over time We extend concepts from the previous FLIPS framework to support streaming FL, requiring new streaming FL approaches that can handle continually evolving models, non-stationary inputs and label distributions. For instance, patient data evolves over time, new symptoms emerge, devices degrade, routines shift, making it essential for ML models to continuously adapt. For example, a wearable device tracking mobility may observe co-variate shifts as patients’ walking patterns change with age or due to temporary conditions like medication side effects. Similarly, label shifts can occur when the prevalence of health outcomes changes, such as an increase in fall risk during winter months or a shift in diagnostic criteria for cognitive decline. Our initial work on ShiftEx Bhope et al. (2025), we developed a streaming FL system designed to detect such distributional changes and adapt accordingly, without compromising patient privacy. ShiftEx uses a Mixture of Experts approach, where different models specialize in distinct patterns, for example, one expert may learn correlations in active seniors with high mobility, while another focuses on sedentary individuals with chronic conditions. Each client periodically analyzes its own data for changes, and the ShiftEx expert assignment technique groups clients with similar shifts to train new experts or reuse the most appropriate expert models. This way, the system builds a dynamic pool of specialized models, balancing adaptation to emerging conditions with the retention of long-term knowledge. Newer approaches like ShiftEx need to be developed to offer a robust, privacy-preserving way for long-term health monitoring and timely interventions, especially in high-stakes environments like senior care where health data stays on-device while still enabling collective intelligence.

5.3 Privacy and security in evolving environments

Data privacy regulations (e.g., GDPR in Europe, CCPA in CA), are increasingly being adopted globally to protect against unsafe data management practices. Such regulations often impose different requirements that depends on the specific data that are handled. This poses several challenges for privacy and security in repurposing existing CPHS infrastructures. Early work by the CareDEX team Chakraborty et al. (2024) explored the role of data privacy issues associated with data related to geriatric patients in senior health facilities. Data-CASE aims to express regulations such as GDPR formally as invariants using a (small set of) data processing concepts that capture system behavior. When such concepts are grounded, i.e., they are provided with a single unambiguous interpretation, systems can achieve compliance by demonstrating that the system-actions they implement maintain the invariants (representing the regulations). Nevertheless, and aggressive optimization of privacy preserving technologies such as homomorphic encryption, secure multi-party computation and federated learning is needed to be able to cope with changing environments and to adapt to applications that might impose different and more strict constraints on data that are handled after the repurposing.

5.4 Scalable end-to-end quality-of-service in the continuum

Next-generation stream processing systems for community-scale IoT applications, such as CareDEX, must handle complex nonfunctional needs, e.g., scalability of input, reliability/timeliness of communication and privacy/security of captured data. In IoT smartspaces, such as an instrumented senior living facility, efficiently batching complex workflows remains challenging in resource-constrained environments. High data rates, combined with real-time processing needs for applications, create the need for efficient edge stream processing techniques. This creates the need for designing scalable edge stream processing workflows in real-world IoT deployments where performance and privacy are key concerns. Furthermore, privacy policy execution/enforcement at the edge for intensive workloads is prohibitively expensive. BATCHIT Wang et al. (2024) is an early effort that leverages intelligent batching techniques to enhance the performance and throughput of streaming in IoT smart spaces. Such middleware approaches are needed to ensure scalable end-to-end Quality-of-Service in the continuum.

5.5 Repurposing of public and private infrastructure

While the scenarios presented previously implicitly apply to known infrastructure in the cloud-edge continuum (i.e., “public infrastructure”), the scope of repurposing and adaptations could also conjoin private infrastructure in an opportunistic manner. By “private infrastructure” we refer to localized, private networked infrastructures that are purpose-built, and walled-in for local use. Examples include: cameras, motion, audio and other sensors affixed to physical infrastructure (walls, poles, rigs, etc.) designed for maintaining security and that sense situational awareness in the immediate vicinity of a home, business, or industrial site; and mobile surveillance artifacts (robots, Unmanned Aerial Vehicles) that monitor physical spaces for anomalous activity. By design, these private infrastructures typically deploy private edge-server or edge-local cloud networks that are walled-in to ensure privacy, minimize information leakage, and prevent malicious actions. These private infrastructures could be opportunistically conjoined with the public infrastructure for the aforementioned scenarios–disaster resilience (CAREDEX), congestion management (UTM systems), and large-scale-visual tracking–to obtain better situational awareness. Several open challenges arise to facilitate such opportunistic public-private infrastructure conjoining, including, but not limited to: interoperability and mechanisms for private discovery/registry; policies and incentives for opting in and sharing the private infrastructure; adaptive and revocable privacy management policies; and data provenance and traceability to guarantee that temporarily shared private data does not persist in the public infrastructure, nor does it flow beyond the temporarily extended perimeter of data sharing.

6 Conclusion

We presented a vision where purpose-built community-scale CPHS could be repurposed seamlessly to achieve new functionality and thereby increase the overall resilience of the commmunity in the face of unexpected events and disasters. Using two motivating community-scale CPHS case studies–NSF project CareDEX and EU project MYRTUS–we outlined the benefits that could accrue from repurposing the computing continuum and the physical infrastructures of these CPHS in extreme and unexpected scenarios such as disasters. We then outlined some of the key enablers required to facilitate repurposing, including containerization, seamless deployment, intelligent orchestration, and security and trust. We closed by summarizing the current status of these repurposing enablers, and presented some of the open research challenges faced in realizing the vision of seamless repurposing of community-scale CPHS to improve overall system resilience.

Finally, we note that human and social factors—including cognitive biases in decision-making within interdependent infrastructures—can significantly affect the outcomes of repurposing strategies. A holistic approach will need to integrate operational as well as organizational aspects Gomez et al. (2019) while applying various optimization Gupta and González, (2026) and modeling approaches Bashir et al. (2026). Although our current work focuses on the technical enablers for infrastructure repurposing, the interplay between human decision-making and system-level adaptations remains an important dimension that warrants further investigation. In particular, the role of modularity as a complementary strategy to cross-integration—balancing the benefits of coordination with the need to contain cascading effects of erroneous decisions—represents a promising direction for future research.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

Edited by: Rosa M. Badia, Barcelona Supercomputing Center, Spain

Reviewed by: César García-Díaz, Pontifical Javeriana University, Colombia

Domenico Talia, University of Calabria, Italy

Footnotes

Authors actively involved in the conception and execution.

Authors actively involved in the conception and execution.

Link:https://www.ospo.noaa.gov/Products/land/hms.html

Link: https://www.airnow.gov