As demand grows, serverless computing systems must scale to meet increasing throughput requirements. Cloud computing easily achieves scalability by allocating abundant and elastic resources. In contrast, edge computing pre-deploys scarce and inelastic resources on site. However, edge applications often need to scale dramatically to handle bursty demand. Because they typically serve fewer users with highly variable workloads, our study shows that peak usage may require up to 8× more resources to be pre-deployed across edge nodes than in a centralized cloud. We observe that a typical serverless request can often be satisfied by different implementations, with dissimilar resource consumption profiles. When an edge application fails to scale up, the culprit is often a single bottleneck resource being fully consumed, with other resources readily available. Motivated by this observation, we introduce SCALEWAVE, a middleware for seamlessly scaling up with different implementations to fully utilize all available resources to achieve scalable serverless computing at the edge. Supporting multiple implementations, however, introduces new challenges for conventional auto-scaler designs. Reactive strategies tend to yield suboptimal performance, while proactive methods struggle with compatibility. SCALEWAVE overcomes these limitations by proactively distributing traffic across implementations, while leveraging existing autoscalers to manage instance scaling reactively for each one. Our evaluations using real workload traces on a heterogeneous cluster of edge devices demonstrate that our design allows edge applications to serve 2x more requests with minimal latency and achieve 50% more successful requests during workload bursts — showcasing substantial gains in scalability and resource efficiency.

Smart devices often use natural language understanding (NLU) to infer user goals. NLU works by extracting user intents from text, then mapping these intents to rigid sets of rules that specify system responses (i.e., slot filling). NLU behaves best when intents and their targets are well-specified (e.g., "set brightness to 50%"), and when the rules implement all desirable responses. NLU falters, however, when responses demand a deeper level of reasoning than predefined rules can provide, e.g., when intents are under-specified (e.g., "make a warm glow") or context-sensitive information is requested (e.g., "why is it so hot in here?"). Toward more flexible user-device interactions, we re-frame the challenge as one of building thoughtful things: devices that leverage a local capacity for reasoning about their internal states to generate responses rather than retrieving them from constrained rule sets. We propose an end-to-end framework for "teaching things to think" that trains a language model to generate valid, device-specific actions and natural language explanations in response to unconstrained language that is out-of-scope for slot filling NLU. Our key insight is that we can bootstrap this generative model from purely synthetic data constructed using a formal definition of a device’s state. The resulting model is small enough to run on-device, with no runtime cloud dependency. We implement two thoughtful things (a lamp and a thermostat) on real hardware and explore their potential to generate and explain device states in response to commands and inquiries.

The urgency in the adoption of the Computing Continuum paradigm, which allows computing services to be de-ployed closer to users, requires a plethora of powerful, distributed devices that host such services. In this context, user-owned devices, such as phones or gaming devices, massively distributed by nature and experiencing a continuous growth in computing resources, are naturally fit to host services, and thus, their incorporation to the Continuum to provide the urgently needed infrastructural support is inevitable. In this future, two key challenges must be addressed: automating service orchestration across a massive number of devices, and ensuring device owners maintain agency over the circumstances and conditions under which services can be hosted on their devices and consumed by other users. As a first step towards this future, we present Atmos, an automated service orchestration platform for integrating user-owned devices in the Computing Continuum. Atmos enforces user-defined policies, automatically adjusting the placement and replication of services across devices. The evaluation of Atmos shows that it enforces all user policies, compared to state-of-the-art service orchestration systems, which violate up to 90.3% of user policies, with minimal impact on the experienced QoS.

Introducing elementary students to computer science and computational thinking (CS/CT) can enhance their problem solving skills and enhance their confidence and sense of belonging in computing. Project moveSMART aims to introduce learning activities into elementary classrooms that address computer science concepts in a way that integrates with core curriculum requirements and promotes physical activity. In this paper, we explore an extension to an initial set of Project moveSMART computer science learning activities to introduce elementary students to machine learning concepts in a way that is integrated with required learning objectives covered in a Physical Education course. Specifically, students use the BBC micro:bit and its on board sensors to capture rhythmic movement data, explore and analyze patterns in the data, and use a learned "dance move recognition" application that uses their data in order to learn about machine learning in an age appropriate way. To demonstrate feasibility of supporting dance move recognition on the resource-constrained device, we developed a prototype, which is able to detect 5 different dance moves with a 96.6% accuracy.

This paper introduces a novel hybrid simulation system (HybriSim) tailored for simulating distributed learning in mobile settings, such as those involving vehicles and pedestrians navigating through cities. Designed to be learning-method independent, the system is compatible with decentralized learning, federated learning, or a combination of the two. It has special relevance for decentralized learning systems that are sensitive to mobility patterns and rely on direct, device-to-device communication. Existing tools for evaluating resource-intensive tasks in opportunistic networks are either purely simulated, which may not accurately reflect system performance, or take the form of testbeds of real devices, which are difficult to scale to use cases involving huge numbers of devices, such as distributed learning. By integrating real devices with virtual simulated devices, HybriSim more accurately mirrors real-world performance and dynamics. This integration not only mitigates the biases associated with pure simulations but also resolves the deployment complexities of conducting simulations entirely on real devices. Our system sets a new benchmark for academic and industry researchers, facilitating more reliable and actionable insights into distributed learning systems in mobility contexts.

In this paper, we define a learning activity for an elementary physical education classroom that simultaneously engages students in physical activity while introducing students to basic principles of reinforcement learning. Reinforcement learning is a sub-domain of machine learning in which an independent agent (in our activity, a student) takes some action or series of actions and receives a reward for the chosen action(s). While reinforcement learning intuitively maps to many activities in our daily lives, our learning activity involves a spy game. Students create sequences of spy moves that generate rewards based on their component moves and the orders in which they are performed. Students then iteratively expand their spy moves in an attempt to receive the maximum reward. The construction of the game will demonstrate that the rewards, while deterministic, do not always follow a greedy pattern, introducing students to basic algorithmic principles. Such an approach that combines physical activity with reinforcement learning connects artificial intelligence education within the broader scope of computing and students' everyday lives.

Opportunistic data sharing allows users to receive real-time, dynamic data directly from peers. These systems not only allow large-scale cooperative sensing but they also empower users to fully control what information is sensed, stored, and shared, enhancing an individual's control over their own potentially private data. While there exist context-aware frameworks that allow individual users to define when and what shared information peers can consume, these approaches have limited expressiveness and do not allow data owners to modulate the granularity of the information released depending on a particular peer or situation. In addition, these frameworks do not consider the consuming peers' privacy, i.e., how much information they have to provide to get access to some desired data. In this paper, we present PADEC, a context-sensitive, privacy-aware framework that allows users to define rich access control rules over their resources and to attach levels of granularity to each rule in order to precisely define who has access to what data when and at what level of detail. Our evaluation shows that PADEC is more expressive than other access control mechanisms and protects the provider’s privacy up to 90% more.

Pervasive computing applications commonly involve user's personal smartphones collecting data to influence application behavior. Applications are often backed by models that learn from the user's experiences to provide personalized and responsive behavior. While models are often pre-trained on massive datasets, federated learning has gained attention for its ability to train globally shared models on users' private data without requiring the users to share their data directly. However, federated learning requires devices to collaborate via a central server, under the assumption that all users desire to learn the same model. We define a new approach, opportunistic federated learning, in which individual devices belonging to different users seek to learn robust models that are personalized to their user's own experiences. However, instead of learning in isolation, these models opportunistically incorporate the learned experiences of other devices they encounter opportunistically. In this paper, we explore the feasibility and limits of such an approach, culminating in a framework that supports encounter-based pairwise collaborative learning. The use of our opportunistic encounter-based learning amplifies the performance of personalized learning while resisting overfitting to encountered data.

With the rapid development and usage of Internet-of-Things (IoT) and smart-home devices, researchers continue efforts to improve the "smartness" of those devices to address daily needs in people's lives. Such efforts usually begin with understanding evolving user behaviors on how humans utilize the devices and what they expect in terms of their behavior. However, while research efforts abound, there is a very limited number of datasets that researchers can use to both understand how people use IoT devices and to evaluate algorithms or systems for smart spaces. In this paper, we collect and characterize more than 50,000 recipes from the online If-This-Then-That (IFTTT) service to understand a seemingly straightforward but complicated question: "What kinds of behaviors do humans expect from their IoT devices?" The dataset we collected contains the basic information of the IFTTT rules, trigger and action events, and how many people are using each rule.

Smartphones have become the perfect companion devices. They have myriad sensors for gathering the context of their owners in order to adapt the behaviour of different applications to the device's situation. This information can also be of great help in enabling the development of social applications that, otherwise, would require a costly and intractable deployment of sensors. Mobile Crowd Sensing systems highly reduce this cost, but realizing this vision using traditional centralized networking primitives requires a constant stream of the sensed data to the cloud in order to store and process it, which in turn leads to the individuals about whom the data is sensed losing control over the privacy of the data. In this paper, we propose an architecture for a device-to-device Mobile Crowd Sensing system and we deepen on a new privacy model that allows users to define access control policies based on their context and the consumer's context.

Advances in mobile computing capabilities and an increasing number of Internet of Things (IoT) devices have enriched the possibilities of the IoT but have also increased the cognitive load required of IoT users. Existing context-aware systems provide various levels of automation in the IoT. Many of these systems adaptively take decisions on how to provide services based on assumptions made a priori. The approaches are difficult to personalize to an individual's dynamic environment, and thus today's smart IoT spaces often demand complex and specialized interactions with the user in order to provide tailored services. We propose rIoT, a framework for seamless and personalized automation of human-device interaction in the IoT. rIoT leverages existing technologies to operate across heterogeneous devices and networks to provide a one-stop solution for device interaction in the IoT. We show how rIoT exploits similarities between contexts and employs a decision-tree like method to adaptively capture a user's preferences from a small number of interactions with the IoT space. We measure the performance of rIoT on two real-world data sets and a real mobile device in terms of accuracy, learning speed, and latency in comparison to two state-of-the-art machine learning algorithms.

The domain of access control has long suffered from a lack of expressiveness in specifying access control policies. Recent approaches have leveraged contextual fingerprinting to formulate access control frameworks for both generating and enforcing access control policies. However, effectively and automatically identifying the context attributes relevant for access has proven challenging and cumbersome. An approach that shows promise in supporting more expressive and easy-to-use attribute-based access control relies on recent advances in continuous neighbor discovery protocols and low cost wireless communication technologies such as Bluetooth Low Energy (BLE). These technologies have created opportunities to build smart environments that can seamlessly and inexpensively provide rich contextual data. These capabilities have the potential to enable new transparent and automatic approaches to defining and evaluating access control policies for mobile users and for detecting anomalous access patterns in smart environments. In this paper, we present the LAD framework that uses raw contextual data available via technologies such as BLE to derive real-time attributes defined by the presence of mobile and static nodes in the nearby environment. Based on user interactions in these environments, our framework learns appropriate access control policies and enforces these policies based on attributes that change in real-time as users move in the smart environment.

We present a demonstration of Jive (Joint Integration of VLC and Encryption), a novel encryption key sharing framework utilizing the emerging wireless technology Visible Light Communication (VLC). Based on the idea of transmitting data by modulating light, we are able to (1) share a secret key within a constrained physical space and (2) leverage this shared key to communicate encrypted information among co-located mobile devices. In this demonstration, we showcase our complete implementation of Jive: a VLC transmitter and a VLC receiver. Both endpoints are built using off-the-shelf components. The VLC link is used to distribute a randomly generated secret key that can only be "observed" by VLC receivers that are physically in the same space as the transmitter. Each receiver is connected over a serial connection to an Android device; we developed applications for Android that take the key from the VLC receiver and subsequently use the key to encrypt or decrypt application data. Our demo invites participants to create their own encrypted messages in the Android application and interact with the VLC prototype as it transmits encryption keys, thus illustrating our system's ability to bootstrap security among physically co-located devices.

This paper investigates a novel encryption key sharing mechanism using the emerging wireless technology Visible Light Communication (VLC). Based on the idea of transmitting data by modulating light, we are able to (1) share a secret key within a constrained physical space and (2) communicate encrypted information among co-located mobile devices using the shared key. We present the demonstration Jive (Joint Integration of VLC and Encryption), a framework to support secret key sharing over Visible Light Communication. In defining Jive, we tackle challenges related to data encoding, message synchronization, and environmental noise to build a reliable, low complexity system using off-the shelf hardware. Our system is capable of sending encryption keys at speeds of more than 750bps using ultra short, high speed light pulses imperceptible to the human eye. Additionally, we have developed an application for Android that interfaces with the VLC device through serial communication so that applications running on mobile devices can subsequently use the keys to encrypt application data. Experimental results illustrate the high accuracy of our system across a variety of different variables. Finally, we position our system for use by a variety of applications that require a high-level of data security among physically co-located devices.

Device-to-device (D2D) communication technologies are growing in availability and popularity and are commonly used to facilitate applications in Internet of Things (IoT) environments. Such environments are characterized by heterogeneous devices, often employing diverse communication technologies with varying energy consumption, discovery ranges, and transmission rates. These complexities pose a daunting setting for the development of IoT applications that could leverage direct communication with proximal mobile and embedded devices. While current approaches focus either on device discovery in the IoT setting or content transfer assuming established communication channels, none facilitate the intelligent discovery of useful devices and the seamless formation of temporary D2D connections to transfer content directly between devices. Our Omni middleware provides both of these features critical in the development of applications that leverage proximal devices in IoT settings. Using Omni, we demonstrate the feasibility of building applications that use heterogeneous D2D communication channels in an efficient and realistic (in terms of energy and time) manner.

Spatiotemporal context is crucial in modern mobile applications that utilize increasing amounts of context to better predict events and user behaviors, requiring rich records of users' or devices' spatiotemporal histories. Maintaining these rich histories requires frequent sampling and indexed storage of spatiotemporal data that pushes the limits of resource-constrained mobile devices. Today's apps offload processing and storing contextual information, but this increases response time, often relies on the user's data connection, and runs the very real risk of revealing sensitive information. In this paper, we motivate the feasibility of on-loading large amounts of context and introduce Programming Abstraction for Contextual On-loading (PACO), an architecture for on-loading data that optimizes for location and time while allowing flexibility in storing additional context. The PACO API's innovations enable on-loading very dense traces of information, even given devices' resource constraints. Using real-world traces and our implementation for Android, we demonstrate that PACO can support expressive application queries entirely on-device. Our quantitative evaluation assesses PACO's energy consumption, execution time, and spatiotemporal query accuracy. Further, PACO facilitates unified contextual reasoning across multiple applications and also supports user-controlled release of contextual data to other devices or the cloud; we demonstrate these assets through a proof-of-concept case study.

Identifying "who is around" is key in a plethora of smart scenarios. While many solutions exist, they often take a theoretical approach, reasoning about protocol behavior with an abstract model that makes simplifying assumptions about the environment. This approach creates a gap between protocol implementations and the models used during design and analysis. In this paper, we take a system approach to continuous neighbor discovery: starting with the concrete technology of Bluetooth Low Energy (BLE) we build a protocol, called BLEnd, tailored to its constraints. Moreover, we also consider the very real effects of packet collisions, to our knowledge a first in this domain. Our ultimate goal is to directly empower developers with the ability to determine the optimal protocol configuration for their applications; in this respect, the slotless operation of BLEnd offers richer alternatives than state-of-the-art protocols. Developers specify the minimum discovery probability, the target discovery latency, and the maximum expected node density; these are used by an optimizer tool to parameterize the BLEnd implementation towards maximum lifetime. This paper shows that BLEnd not only achieves the user-specified goals, but does so more efficiently than analogous configurations of competing protocols.

In Cyber-Physical Systems (CPS), cyber and physical components must work seamlessly in tandem. Runtime verification of CPS is essential yet very difficult, due to deployment environments that are expensive, dangerous, or simply impossible to use for verification tasks. A key enabling factor of runtime verification of CPS is the ability to integrate real-time simulations of portions of the CPS into live running systems. We propose a verification approach that allows CPS application developers to opportunistically leverage real-time simulation to support runtime verification. Our approach, termed B race B ind , allows selecting, at runtime, between actual physical processes or simulations of them to support a running CPS application. To build B race B ind , we create a real-time simulation architecture to generate and manage multiple real-time simulation environments based on existing simulation models in a manner that ensures sufficient accuracy for verifying a CPS application. Specifically, B race B ind aims to both improve simulation speed and minimize latency, thereby making it feasible to integrate simulations of physical processes into the running CPS application. B race B ind then integrates this real-time simulation architecture with an existing runtime verification approach that has low computational overhead and high accuracy. This integration uses an aspect-oriented adapter architecture that connects the variables in the cyber portion of the CPS application with either sensors and actuators in the physical world or the automatically generated real-time simulation. Our experimental results show that, with a negligible performance penalty, our approach is both efficient and effective in detecting program errors that are otherwise only detectable in a physical deployment.

Supporting mobile applications in densely populated environments requires connecting mobile users and their devices with the surrounding digital landscape. Device-to-device communications (without the support of an infrastructure) will play a critical role in facilitating transparent access to proximate digital resources. A variety of approaches exist to support device-to-device dissemination, yet very few capitalize on the contextual history of disseminated data itself to distribute additional data. We introduce a fully decentralized model that captures the causal history of shared information across a lifetime of device-to-device propagation. Our approach enhances each transmitted message with a spatiotemporal trajectory, a portion of the message's contextual history, which provides a window into the time-varying state of the network and the overall spreading behavior of the message. We benchmark the performance of spatiotemporal trajectories at a global level and demonstrate the practical utility of our constructs for making intelligent distributed routing decisions through two use cases. The first illustrates how locally-available contextual history may be used to make inferences about remote nodes' knowledge and thus direct routing of future messages. The second example uses trajectories from different types of data to identify commonly co-located data and devices, which can then be used as substitutable routing targets.

Acquiring local context information and sharing it among co-located devices is critical for emerging pervasive computing applications. The devices belonging to a group of co-located people may need to detect a shared activity (e.g., a meeting) to adapt their devices to support the activity. Today's devices are almost universally equipped with device-to-device communication that easily enables direct context sharing. While existing context sharing models tend not to consider devices' resource limitations or users' constraints, enabling devices to directly share context has significant benefits for efficiency, cost, and privacy. However, as we demonstrate quantitatively, when devices share context via device-to-device communication, it needs to be represented in a size-efficient way that does not sacrifice its expressiveness or accuracy. We present CHITCHAT, a suite of context representations that allows application developers to tune tradeoffs between the size of the representation, the flexibility of the application to update context information, the energy required to create and share context, and the quality of the information shared. We can substantially reduce the size of context representation (thereby reducing applications' overheads when they share their contexts with one another) with only a minimal reduction in the quality of shared contexts.

In pervasive computing environments, understanding the context of an entity is essential for adapting the application behavior to changing situations. In our view, context is a high-level representation of a user or entity's state and can capture location, activities, social relationships, capabilities, etc. Inherently, however, these high-level context metrics are difficult to capture using uni-modal sensors only and must therefore be inferred using multi-modal sensors. A key challenge in supporting context-aware pervasive computing is how to determine multiple high-level context metrics simultaneously and energy-efficiently using low-level sensor data streams collected from the environment and the entities present therein. A key challenge is addressing the fact that the algorithms that determine different high-level context metrics may compete for access to low-level sensors. In this paper, we first highlight the complexities of determining multiple context metrics as compared to a single context and then develop a novel framework and practical implementation for this problem. The proposed framework captures the tradeoff between the accuracy of estimating multiple context metrics and the overhead incurred in acquiring the necessary sensor data streams. In particular, we develop two variants of a heuristic algorithm for multi-context search that compute the optimal set of sensors contributing to the multi-context determination as well as the associated parameters of the sensing tasks (e.g., the frequency of data acquisition). Our goal is to satisfy the application requirements for a specified accuracy at a minimum cost. We compare the performance of our heuristics with a brute-force based approach for multi-context determination. Experimental results with SunSPOT, Shimmer and Smartphone sensors in smart home environments demonstrate the potential impact of the proposed framework.

Cyber-Physical Systems (CPS) have gained wide popularity, however, developing and debugging CPS remain significant challenges. Many bugs are detectable only at runtime under deployment conditions that may be unpredictable or at least unexpected at development time. The current state of the practice of debugging CPS is generally ad hoc, involving trial and error in a real deployment. For increased rigor, it is appealing to bring formal methods to CPS verification. However developers often eschew formal approaches due to complexity and lack of efficiency. This paper presents Brace Assertion, a specification framework based on natural language queries that are automatically converted to a determinitic class of timed automata used for runtime monitoring. To reduce runtime overhead and support properties that reference predicate logic, we use a second monitor automaton to create filtered traces on which to run the analysis using the specification monitor. We evaluate the Brace Assertion framework using a real CPS case study and show that the framework is able to minimize runtime overhead with an increasing number of monitors.

Sensing applications often require participants to share context information about the physical social, or network environment in which they operate. Building shared views of context requires exchanging sensed information, often via peer-to-peer links. In-network aggregation enables efficient distributed data collection, but the goal has been almost exclusively collect a single aggregate value at a single sink node. In contrast, we design and implement a simple protocol for exchanging context information in aggregate in a peer-to-peer fashion, where every node needs to acquire a shared view of the aggregate context. In our protocol, when a node receives new context information from a neighboring node, it aggregates the new information into its local view of the shared state of the world which it then subsequently shares with its neighbors. We demonstrate (both theoretically and empirically) the situations in which participants' raw context information is fully or partially recoverable by other participants from an aggregate and quantify the tradeoffs in communication overhead for the quality of shared context knowledge. Compared with non-aggregation communication for sharing context values among 100 nodes in a simulated network, we show an overhead savings of at least 78.0%, and an overhead savings of 66.0% with 99.8% average accuracy in a 54 node emulated network driven by real world data.

Mobile computing increasingly often entails applications that embody opportunistic or delay-tolerant communication, and while much work has focused on refining and optimizing the technical underpinnings for providing delay-tolerant communication constructs, there is almost a complete lack of support for integrating opportunistic communication functionality at the application level. This paper introduces MadApp, an application-level development framework that provides tailored abstractions and support infrastructure for creating dynamic web pages that can incorporate received content from various opportunistic communication channels on-the-fly. We describe multiple application scenarios in which these constructs can seamlessly apply, and provide a complete conceptual and concrete architecture and implementation for MadApp. We evaluate MadApp's support for opportunistic mobile computing web applications using two different mobility trace data sets collected from the real-world. This paper demonstrates that MadApp enables opportunistic mobile computing applications to begin to leverage the significant advances in delay-tolerant communication research, opening doors for even more dynamic and adaptive applications.

The vision of the Internet of Things (IoT) will enable networked environments populated with vast amounts of data that can be exploited by humans. The volume of digitally available data in such emerging computing spaces presents an imminent need for search mechanisms that enable humans and applications to find relevant information within their digitally accessible physical surroundings. This paper presents Gander, a search engine for these pervasive computing spaces enabled by the IoT and characterized by large volumes of highly transient data. Gander is founded on a novel conceptual model of search that resolves queries about a user's here and now by leveraging proximally available resources in the here and now. We formally describe the model underlying Gander, describe the networking protocols that enable Gander's search, and provide a realization of Gander via an extensible framework. Employing this Gander framework, we describe a concrete middleware implementation for wirelessly networked environments. We evaluate this implementation of Gander through a user study that examines the perceived utility of myGander, a real-world mobile application enabled by the Gander middleware, and we benchmark the performance of Gander in large pervasive computing spaces through network simulation.

We introduce SESENA 2013, the fourth in a series of workshops devoted to software engineering for sensor network applications. The workshop took place in San Francisco (USA) on May 21, 2013, in conjunction with the 35thACM/IEEE International Conference on Software Engineering (ICSE). The goal was to bring together research from both the field of software engineering (SE) and wireless sensor networks (WSN) to engender exchange and discussion on shared research goals and agendas. More information can be found on the workshop website http://www.sesena.info.

Multi-robot patrol is a fundamental application of multi-robot systems. While much theoretical work exists providing an understanding of the optimal patrol strategy for teams of coordinated homogeneous robots, little work exists on building and evaluating the performance of such systems for real. In this paper, we evaluate the performance of multirobot patrol in a practical outdoor distributed robotic system, and evaluate the effect of different coordination schemes on the performance of the robotic team. The multi-robot patrol algorithms evaluated vary in the level of robot coordination: no coordination, loose coordination, and tight coordination. In addition, we evaluate versions of these algorithms that distribute state information-either individual state, or entire team state (global-view state). Our experiments show that while tight coordination is theoretically optimal, it is not practical in practice. Instead, uncoordinated patrol performs best in terms of average waypoint visitation frequency, though loosely coordinated patrol that shares only individual state performed best in terms of worst-case frequency. Both are significantly better than a loosely coordinated algorithm based on sharing global-view state. We respond to this discrepancy between theory and practice, caused primarily by robot heterogeneity, by extending the theory to account for such heterogeneity, and find that the new theory accounts for the empirical results.

Developing cyber-physical systems (CPS) is challenging because correctness depends on both logical and physical states, which are collectively difficult to observe. The developer often need to repeatedly rerun the system while observing its behavior and tweak the hardware and software until it meets minimum requirements. This process is tedious, error-prone, and lacks rigor. To address this, we propose BRACE, A framework that simplifies the process by enabling developers to correlate cyber (i.e., logical) and physical properties of the system via assertions. This paper presents our initial investigation into the requirements and semantics of such assertions, which we call CPS assertions. We discusses our experience implementing and using the framework with a mobile robot, and highlight key future research challenges.

Autonomous intersection management (AIM) is a new intersection control protocol that exploits the capabilities of autonomous vehicles to control traffic at intersections in a way better than traffic signals and stop signs. A key assumption of this protocol is that vehicles can always follow their trajectories. But mechanical failures can occur in real life, causing vehicles to deviate from their trajectories. A previous approach for handling mechanical failure was to prevent vehicles from entering the intersection after the failure. However, this approach cannot prevent collisions among vehicles already in the intersection or too close to stop because (1) the lack of coordination among vehicles can cause collisions during the execution of evasive actions; and (2) the intersection may not have enough room for evasive actions. In this paper, we propose a preemptive approach that pre-computes evasion plans for several common types of mechanical failures before vehicles enter an intersection. This preemptive approach is necessary because there are situations in which vehicles cannot evade without pre-allocation of space for evasion. We present a modified AIM protocol and demonstrate the effectiveness of evasion plan execution on a miniature autonomous intersection testbed.

Pervasive computing applications often involve sensor-rich networking environments that capture various types of user contexts such as locations, activities, vital signs, and so on. Such context information is useful in a variety of applications, for example, monitoring health information to promote independent living in "aging-in-place” scenarios, or providing safety and security of people and infrastructures. In reality, both sensed and interpreted contexts are often ambiguous, thus leading to potentially dangerous decisions if not properly handled. Therefore, a significant challenge in the design and development of realistic and deployable context-aware services for pervasive computing applications lies in the ability to deal with ambiguous contexts. In this paper, we propose a resource-optimized, quality-assured context mediation framework for sensor networks. The underlying approach is based on efficient context-aware data fusion, information-theoretic reasoning, and selection of sensor parameters, leading to an optimal state estimation. In particular, we apply dynamic Bayesian networks to derive context and deal with context ambiguity or error in a probabilistic manner. Experimental results using SunSPOT sensors demonstrate the promise of this approach.

In pervasive computing environments, understanding the context of an entity is essential for adapting the application behavior to changing situations. In our view, context is a high-level representation of a user or entity's state and can capture location, activities, social relationships, capabilities, etc. Inherently, however, these high-level context metrics are difficult to capture using uni-modal sensors only, and must therefore be inferred with the help of multi-modal sensors. However a key challenge in supporting context-aware pervasive computing environments, is how to determine in an energy-efficient manner multiple (potentially competing) high-level context metrics simultaneously using low-level sensor data streams about the environment and the entities present therein. In this paper, we first highlight the intricacies of determining multiple context metrics as compared to a single context, and then develop a novel framework and practical implementation for this problem. The proposed framework captures the tradeoff between the accuracy of estimating multiple context metrics and the overhead incurred in acquiring the necessary sensor data stream. In particular, we develop a multi-context search heuristic algorithm that computes the optimal set of sensors contributing to the multi-context determination as well as the associated parameters of the sensing tasks. Our goal is to satisfy the application requirements for a specified accuracy at a minimum cost. We compare the performance of our heuristic based framework with a brute-forced approach for multi-context determination. Experimental results with SunSPOT sensors demonstrate the potential impact of the proposed framework.

The number of different types of devices on the home network is expanding rapidly. While this explosion of innovation provides compelling new devices to consumers, there are challenges ensuring compatibility among these devices and providing a comprehensive user interface that supports consumers managing their digital content across several devices. The Digital Living Network Alliance (DLNA) has specified an architecture that enables interoperability between the various devices and allows a user to enjoy the desired content across several devices. To complement DLNA, we investigate ontological representation from semantic web technology to model the interaction between multiple home devices and to provide an enriched and more comprehensive metadata based multimedia search. This removes the user burden of searching each device individually. Development of a prototype incorporating these ideas has begun, using a well known semantic web toolkit Jena.

Mobile ad hoc networks (MANETs) can have widely varying characteristics under different deployments, and previous studies show that the characteristics impact the behavior of routing protocols for MANETs. To deploy applications successfully in MANETs, application developers need to comprehend the potential behavior of any underlying protocol used. In mobile networks, a major component of many of these routing protocols is some form of flooding, which facilitates message delivery over an entire network in a relatively reliable way. Several MANET protocols use flooding to support the distribution of route request messages as well as delivering broadcast packets. Therefore, to developers of applications for MANETs, a major task in understanding MANET protocols is estimating the performance of flooding in a given operating environment. In this work, we develop an analytical model for the delay experienced in flooding a message. We model the one hop delay of a flooding message in MANETs in terms of parameters that can be acquired either from a system configuration or from application designers.

Many pervasive computing applications continuously monitor state changes in the environment by acquiring, interpreting and responding to information from sensors embedded in the environment. However, it is extremely difficult and expensive to obtain a continuous, complete, and consistent picture of a continuously evolving operating environment. One standard technique to mitigate this problem is to employ mathematical models that compute missing data from sampled observations thereby approximating a continuous and complete stream of information. However, existing models have traditionally not incorporated a notion of temporal validity, or the quantification of imprecision associated with inferring data values from past or future observations. In this paper, we support continuous monitoring of dynamic pervasive computing phenomena through the use of a series of snapshot queries. We define a decay function and a set of inference approaches to filling in missing and uncertain data in this continuous query.We evaluate the usefulness of this abstraction in its application to complex spatio-temporal pattern queries in pervasive computing networks.

Queries are convenient abstractions for the discovery of information and services, as they offer content-based information access. In distributed settings, query semantics are well-defined, for example, queries are often designed to satisfy ACID transactional properties. When query processing is introduced in a dynamic network setting, achieving transactional semantics becomes complex due to the open and unpredictable environment. In this article, we propose a query processing model for mobile ad hoc and sensor networks that is suitable for expressing a wide range of query semantics; the semantics differ in the degree of consistency with which query results reflect the state of the environment during query execution. We introduce several distinct notions of consistency and formally express them in our model. A practical and significant contribution of this article is a protocol for query processing that automatically assesses and adaptively provides an achievable degree of consistency given the operational environment throughout its execution. The protocol attaches an assessment of the achieved guarantee to returned query results, allowing precise reasoning about a query with a range of possible semantics. We evaluate the performance of this protocol and demonstrate the benefits accrued to applications through examples drawn from an industrial application.

Ubiquitous (or smart) healthcare applications envision sensor rich computing and networking environments that can capture various types of contexts of patients (or inhabitants of the environment), such as their location, activities and vital signs. Such context information is useful in providing hea

In pervasive computing environments, applications find themselves in constantly changing operating conditions. Such applications often need to discover locally available resources on-demand. Communication protocols have been developed that base discovery not on the unique address of the destination but on application-level characteristics of the destination host. Previous work has focused almost exclusively on purely on-demand protocols to achieve these resource connections. However, because the types of resources desired may be common across applications, the discovery and routing tasks can benefit from some degree of proactivity. In this paper, we describe our adaptive approach to incorporating resource advertisement in an application-driven routing protocol. We describe the adaptation mechanism in our protocol that allows the proactive component to dynamically tune its behavior to operating conditions.

In wireless sensor network applications, the potential to use cooperation to resolve user queries remains largely untapped. Efficiently answering a user's questions requires identifying the correct set of nodes that can answer the question and enabling coordination between them. In this article, we propose a query domain abstraction that allows an application to dynamically specify the nodes best suited to answering a particular query. Selecting the ideal set of heterogeneous sensors entails answering two fundamental questions— how are the selected sensors related to one another, and where should the resulting sensor coalition be located. We introduce two abstractions, the proximity function and the reference function , to precisely specify each of these concerns within a query. All nodes in the query domain must satisfy any provided proximity function, a user-defined function that constrains the relative relationship among the group of nodes (e.g., based on a property of the network or physical environment or on logical properties of the nodes). The selected set of nodes must also satisfy any provided reference function, a mechanism to scope the location of the query domain to a specified area of interest (e.g., within a certain distance from a specified reference point). In this article, we model these abstractions and present a set of protocols that accomplish this task with varying degrees of correctness. We evaluate their performance through simulation and highlight the tradeoffs between protocol overhead and correctness.

This paper argues for explicit consideration of data fidelity during development of context-aware systems. Increasing the amount of data captured, stored, and distributed does not always translate into into increased fidelity, and we can leverage this to avoid unnecessary energy overhead and increase device battery life. We introduce Context-Awareness Fidelity Expression (CAFE) as a framework for deliberately specifying application data fidelity adaptation with the aim of using context-awareness to reduce energy consumption.

Context-aware computing is characterized by the software's ability to continuously adapt its behavior to an environment over which it has little control. This style of interaction is imperative in ad hoc mobile networks that consist of numerous mobile hosts coordinating opportunistically via transient wireless connections. In this paper, we provide a formal abstract characterization of an application's context that extends to encompass a neighborhood within the ad hoc network. We provide a context specification mechanism that allows individual applications to tailor their operating contexts to their personalized needs. We describe a context maintenance protocol that provides this context abstraction in ad hoc networks through continuous evaluation of the context. This relieves the application developer of the obligation of explicitly managing mobility and its implications on behavior. We also characterize the performance of this protocol in ad hoc networks through simulation experiments. Finally, we examine real-world application examples demonstrating its use.

Efficient query processing in sensor networks involves identifying groups of nodes that coordinate to satisfy applications' requests. In this paper, we propose a query domain abstraction that allows an application to dynamically specify the nodes best suited to answering a particular query. To self-organize into such a coalition, nodes must satisfy a proximity function, a user-defined function that constrains the relative relationship among the group of nodes (e.g., based on a property of the network or physical environment or a logical property of the nodes). The proximity function removes the need to explicitly tag nodes with context information, and it provides a convenient mechanism for forming coalitions on-the-fly at query time. This facilitates the deployment of general-purpose sensor networks, where multiple applications can be run in the same network at the same time. In this paper, we model this abstraction, present a protocol to support the abstraction, and evaluate their performance.

As sensor networks are increasingly used to support pervasive computing, we envision an instrumented environment that can provide varying amounts of information to mobile applications immersed within the network. Such a scenario deviates from existing deployments of sensor networks which are often highly application-specific and funnel information to a central collection point. We instead target scenarios in which multiple mobile applications will leverage sensor network nodes opportunistically and unpredictably. Such situations require new communication abstractions that enable immersed devices to interact directly with available sensors, reducing both communication overhead and data latency. This paper introduces scenes, which applications create based on their communication requirements, abstract properties of the underlying network communication, and properties of the physical environment. This paper reports on the communication model, an initial implementation, and its performance in varying scenarios

Recently, embedded sensor usage has increased thanks to the proliferation of hardware and software addressing resource constraints. While the increased usage is a good start, it is important to adopt good software engineering principles early for many reasons: simplified programming and deployment, improved reuse, and added efficiency. This poster presents SASSI, the Sliverware Architecture for Sensor System Integration, which provides a unique embedded sensor application architecture.

Queries are convenient abstractions for the discovery of information and services, as they offer content-based information access. In distributed settings, query semantics are well-defined, e.g., they often satisfy ACID transactional properties. In a dynamic network setting, however, achieving transactional semantics becomes complex due to the openness and unpredictability. In this paper, we propose a query processing model for mobile ad hoc and sensor networks suitable for expressing a wide range of query semantics; the semantics differ in the degree of consistency with which results reflect the state of the environment during execution. We introduce several distinct notions of consistency and formalize them. A practical contribution of this paper is a protocol for query processing that automatically assesses and adaptively provides an achievable degree of consistency given the state of the operational environment throughout its execution. The protocol attaches an assessment of the achieved guarantee to returned query results, allowing precise reasoning about a query with a range of possible semantics.

This paper describes application demonstrations of a new middleware that supports dynamic decision support over networks of resource-constrained devices. For the purposes of our demonstrations, we tap into intelligent job site applications for the construction domain. The demonstrations are carefully constructed to highlight our middleware's ability to provide on-demand access to local data, aggregation of data across dynamically defined regions, fusion of heterogeneous sensor data, and intelligent application-sensitive sensor clustering. The paper briefly describes the middleware and the details of the demonstrations

This work addresses the need for application-aware adaptive communication in mobile ad hoc networks that creates network routes based on applications' dynamic resource requests. We introduce an intuitive generalization to source routing which facilitates discovery of a resource in a mobile ad hoc network and the creation and maintenance of a route from the requesting host to the discovered destination. We thus eliminate the requirement that existing routing protocols be coupled with a name or resource resolution protocol, instead favoring an entirely reactive approach to accommodate significant degrees of mobility and uncertainty. We also present a performance evaluation and a comparison to existing alternatives

Sensor networks are becoming increasingly pervasive. Existing methods of aggregation in sensor networks offer mostly standard mathematical operators over homogeneous data types. In this paper, we instead focus on supporting emerging scenarios in which applications will need to extract abstracted measurements from diverse sets of sensor network nodes. This paper introduces the virtual sensors abstraction that enables an application developer to programmatically specify an application's high-level data requirements. This paper reports on our initial work with the virtual sensors and the results of our prototype implementation.

Today's mobile applications require constant adaptation to their changing environments, or contexts. Technological advances have increased the pervasiveness of mobile computing devices such as laptops, handhelds, and embedded sensors. The sheer amount of context information available for adaptation places a heightened burden on application developers as they must manage and utilize vast amounts of data from diverse sources. Facilitating programming in this data-rich environment requires a middleware that provides context information to applications in an abstract form. In this paper, we demonstrate the feasibility of such a middleware that allows programmers to focus on high-level interactions among programs and to employ declarative abstract context specifications in settings that exhibit transient interactions with opportunistically encountered components. We also discuss the novel context-aware abstractions the middleware provides and the programming knowledge necessary to write applications using it. Finally, we provide examples demonstrating the infrastructure's ability to support differing tasks from a wide variety of application domains

The design of ad hoc mobile applications often requires the availability of a consistent view of the application state among the participating hosts. Such views are important because they simplify both the programming and verification tasks. We argue that preventing the occurrence of unannounced disconnection is essential to constructing and maintaining a consistent view in the ad hoc mobile environment. In this light, we provide the specification for a partitionable group membership service supporting ad hoc mobile applications and propose a protocol for implementing the service. A unique property of this partitionable group membership is that messages sent between group members are guaranteed to be delivered successfully, given appropriate system assumptions. This property is preserved over time despite movement and frequent disconnections. The protocol splits and merges groups and maintains a logical connectivity graph based on a notion of safe distance. An implementation of the protocol in Java is available for testing. This work is used in an implementation of LIME, a middleware for mobility that supports transparent sharing of data in both wired and ad hoc wireless environments.

Context-aware computing is characterized by the ability of a software system to continuously adapt its behavior to a changing environment over which it has little or no control. Previous work along these lines presumed a rather narrow definition of context, one that was centered on resources immediately available to the component in question, e.g., communication bandwidth, physical location, etc. This paper explores context-aware computing in the setting of ad hoc networks consisting of numerous mobile hosts that interact with each other opportunistically via transient wireless interconnections. We extend the context to encompass awareness of an entire neighborhood within the ad hoc network. A formal abstract characterization of this new perspective is proposed. The result is a specification method and associated context maintenance protocol. The former enables an application to define an individualized context, one that extends across multiple mobile hosts in the ad hoc network. The latter makes it possible to delegate the continuous reevaluation of the context and the performance of operations on it to some middleware operating below the application level. This relieves application development of the obligation of explicitly managing mobility and its implications on the component's behavior.

Some of the most dynamic systems being built today consist of physically mobile hosts and logically mobile agents. Such systems exhibit frequent configuration changes and a great deal of resource variability. Applications executing under these circumstances need to react continuously and rapidly to changes in operating conditions and must adapt their behavior accordingly. The development of such applications demands a reexamination of the notion of context and the mechanisms used to manage the application's response to contextual changes. This paper introduces EgoSpaces, a coordination model and middleware for ad hoc mobile environments. EgoSpaces focuses on the needs of application development in ad hoc environments by proposing an agent-centered notion of context, called a view, whose scope extends beyond the local host to data and resources associated with hosts and agents within a subnet surrounding the agent of interest. An agent may operate over multiple views whose definitions may change over time. An agent uses declarative specifications to constrain the contents of each view by employing a rich set of constraints that take into consideration properties of the individual data items, the agents that own them, the hosts on which the agents reside, and the physical and logical topology of the ad hoc network. This paper formalizes the concept of view, explores the notion of programming against views, discusses possible implementation strategies for transparent context maintenance, and describes our current prototype of the system. We include examples to illustrate the expressive power of the view abstraction and to relate it to other research on coordination models and middleware.