This paper addresses the failure detection and recovery problem in visual-inertial based Simultaneous Localization and Mapping (SLAM) systems for large-scale indoor environments. Camera and Inertial Measurement Unit (IMU) are popular choices for SLAM in many robotics tasks (e.g., navigation) due to their complementary sensing capabilities and low cost. However, vision has inherent challenges even in well-lit scenes, including motion blur, lack of features, or even accidental camera blockage. These failures can cause drifts to accumulate over time and can severely impact the scalability of existing solutions to large areas. To address these issues, we propose an automatic map generation service with (i) a failure detection method based on visual feature tracking quality using a health tracker which identifies and discards faulty measurements and (ii) a continuous session merging approach in SLAM. Taken together, this allows us to handle erroneous data without any manual intervention, and allows us to scale to extremely large spaces. The proposed system has been validated on benchmark datasets. Also, experimental results on multiple custom large-scale grocery stores, each between 1700 m2to 3700 m2, and duration 60 to 80 minutes, are presented. Our approach shows the lowest error in all large-scale SLAM cases when compared with state-of-the-art visual-inertial SLAM packages, which often produce highly erroneous trajectories or lose track. Additionally, we provide dense 3D reconstruction with the presence of a depth camera by simply registering the point cloud from RGB-D image with respect to the SLAM generated trajectory – and the quality of the reconstruction illustrates the efficacy of our proposed method.

The focus of this paper is on designing a robust information dissemination protocol for Vehicular Ad-Hoc Networks (VANETs) that works across a broad range of application scenarios by dynamically adapting its parameters and transmission strategy to specific information requirements and traffic conditions. Applications of V2V communication can be broadly classified as safety, navigation and environmental efficiency. Safety applications require high frequency and high precision messages at short range (500m) within a single communication hop, navigation applications need low frequency and lower precision information at medium range (1500m/1 mile) and long range aggregate information (10000m/6 miles) can be used for enhancing environmental efficiency and routing. While network protocols for each of these applications have been individually studied, there is scope to study their behavior when they coexist in the network. It is important that the quality of one hop safety messages is not sacrificed when long range communication is introduced. RoAdNet enables long range information dissemination, while maintaining high data.

Environmental monitoring applications often require 24/7 operation in harsh, low resource (e.g. power and communication) environments over a large scale area with ad-hoc deployment of sensors. Data processing at the sensor is required to minimize communication overhead. Such an application scenario presents opportunities for research in wireless sensor networks (WSN)s that are distinct from existing commercial off-the-shelf (COTS) solutions. We present a novel modular, highly flexible, hardware solution with a core feature of a System on a Chip (SoC) with add-ons such as memories, interfaces, and different transmission input/output I/O modalities. The system can manage, process, and transmit data directly within an ad-hoc self healing, self forming, mesh network over long distance (19 km between nodes in the current implementation) or as a stand-alone system. Hardware has been produced and the system has been validated in real-world deployments.

To meet the demanding requirements in the growing area of wireless sensing applications, some sensing platforms have included low-power application-specific hardware to process the sensor data for compression and pre-classification of the relevant information. While this additional hardware can reduce the overall power consumption of the system, a unique hardware solution is required for each application. To diminish this burden, we will demonstrate a reconfigurable analog/mixed-signal sensing platform. At the hardware-level, this platform consists of a reconfigurable integrated circuit containing many commonly used circuit components that can be connected in any configuration to perform sensor interfacing and ultra-low-power signal processing. At the software level, this platform provides a framework for abstracting the underlying hardware. We will demonstrate how our platform allows a developer to create applications ranging from standard sensor interfacing techniques to more complicated intelligent pre-processing and wake-up detection, without the necessity of circuit-level expertise.

The requirements of many wireless sensing applications approach, or even exceed, the limited hardware capabilities of energy-constrained sensing platforms. To achieve such demanding requirements, some sensing platforms have included low-power application-specific hardware---at the expense of generality---to pre-process the sensor data for reduction to only the relevant information. While this additional hardware can save power by reducing the activity of the microcontroller and radio, a unique hardware solution is required for each application, which presents an unrealistic burden in terms of design time, cost, and ease of integration. To diminish these burdens, we present a reconfigurable analog/mixed-signal sensing platform in this work. At the hardware-level, this platform consists of a reconfigurable integrated circuit containing many commonly used signal-processing blocks and circuit components that can be connected in any configuration. At the software level, this platform provides a framework for abstracting this underlying hardware. We demonstrate how to quickly develop new applications on this platform, ranging from standard sensor interfacing techniques to more complicated intelligent pre-processing and wake-up detection. We also demonstrate how to integrate this platform with commonly used wireless sensor nodes and embedded-system platforms.

The limited power budgets of sensor networks necessitate some level of in-network pre-processing to reduce communication overhead. The low power consumption of analog signal processing (ASP) is well-suited for this task. However, the quick adoption of this technology has been restrained by the fact that ASP implementation requires a priori knowledge of the application space. Our solution to this challenge is to enable run-time reconfiguration through the use of a field-programmable analog array (FPAA). In the same way that reconfigurable digital systems allow system designers to change the infrastructure of digital blocks, an FPAA allows an application developer to change the infrastructure of, and even tune, ASP blocks without circuit-level expertise. We will demonstrate that an FPAA can be used to (1) facilitate the use of ASP to reduce power consumption, and to (2) allow run-time reconfigurability to maximize ASP impact.

In this paper, we describe VCAST, an algorithm for obtaining individual vehicle location as well as aggregate traffic information over an infrastructure-less multi-hop wireless vehicular network. VCAST can be used to improve safety against collisions, inform about approaching emergency vehicles and lane merging vehicles (that may even be beyond a single hop communication range) and to enable dynamic routing and navigation techniques by providing aggregate traffic information in an extended neighborhood. To ensure scalability in forwarding information over multiple hops, we exploit a notion of distance sensitivity in information propagation, by which traffic information is propagated at a rate that decreases linearly with distance from the source. By doing so, traffic information can be obtained with a staleness, which is a measure of error in the traffic information, that is bounded by O(dh2) where dhis the communication hop distance from the source of the information. At the same time, the required communication rate per unit time at each node only depends on the radius of the region in terms of communication hops, and not on the vehicular density or the number of vehicles in the region. VCAST does not require any special hardware or modification to vehicular transmission standards; instead it can simply piggyback on basic Here I am communication for vehicular networks.

In this paper, we discuss a low power embedded sensor node architecture we are developing for distributed sensor network systems deployed in a natural environment. In particular, we examine the sensor node for energy efficient processing-at-the-sensor. We analyze the following modes of operation; event detection, data acquisition, and data processing using low power, high performance embedded technology such as specialized embedded DSP processors and low power FPGAs at the sensing node. We use compute intensive sensor node applications: an acoustic vehicle classifier (frequency domain analysis) and a video license plate identification application (learning algorithm). We report performance and energy for these applications and discuss the system architecture design trade offs.

In-network processing is recommended for many sensor network applications to reduce communication and improve energy efficiency. However, constraints on memory, speed, and energy currently limit the processing capabilities within a sensor network. In this paper, we describe how ultra-low-power analog circuitry can be integrated with sensor nodes to create energy-efficient sensor networks. We present a custom analog front-end which performs spectral analysis at a fraction of the power used by a digital counterpart. We then show that the front-end can be combined with existing sensor nodes to (1) selectively wake up the mote based upon spectral content of the signal, thus increasing battery life without missing interesting events, and to (2) achieve low-power signal analysis using an analog spectral decomposition block, freeing up digital computation resources for higher-level analysis.

In-network processing is recommended for many sensor network applications to reduce communication and improve energy efficiency. However, constraints on memory, speed, and energy currently limit the processing capabilities within a sensor network. By integrating ultra-low-power analog circuitry with sensor nodes, we can reduce the node's power consumption and extend the node's processing capacity. We present a custom analog front-end which performs spectral analysis at a fraction of the power used by a digital counterpart. This front-end has been combined with a TelosB mote to (1) selectively wake up the mote based upon spectral content of the signal, thus increasing battery life without missing interesting events, and to (2) achieve low-power signal analysis using an analog spectral decomposition block, freeing up digital computation resources for higher-level analysis.

Distributed observation and control of mobile objects via static wireless sensors demands timely information in a distance-sensitive manner: Information about closer objects is required more often and more quickly than that of farther objects. In this article, we present a wireless sensor network protocol, Trail, that supports distance-sensitive tracking of mobile objects for in-network subscribers upon demand. Trail achieves a find time that is linear in the distance from a subscriber to an object, via a distributed data structure that is updated only locally when the object moves. Notably, Trail does not partition the network into a hierarchy of clusters and clusterheads, and as a result Trail has lower maintenance costs, is more locally fault tolerant, and it better utilizes the network in terms of load balancing and minimizing the size of the data structure needed for tracking. Moreover, Trail is reliable and energy efficient, despite the network dynamics that are typical of wireless sensor networks. Trail can be refined by tuning certain parameters, thereby yielding a family of protocols that are suited for different application settings such as rate of queries, rate of updates, and network size. We evaluate the performance of Trail by analysis, simulations in a 90 × 90 sensor network, and experiments on 105 Mica2 nodes in the context of a pursuer-evader control application.

Surveillance systems based on sensor network technology have been shown to successfully detect, classify and track targets of interest over a large area. State information collected via the sensor network also enables these systems to actuate mobile agents so as to achieve surveillance goals such as target capture and asset protection. But satisfying these goals is complicated by the fact that track information in a sensor network is routed to mobile agents through multi-hop communication links and is thus subject to delays and losses. In addition, as the sensor network is scaled in size, high throughput rates for all pursuers cannot be sustained at all times, which necessitates a network communication strategy that adapts to pursuer information requirements.In this paper, we concentrate on the formulation of optimal pursuit control strategies in the presence of network effects, assuming that target track information has been established locally in the sensor network. We adapt ideas from the theory of differential games to networked games --including ones involving non-periodic track updates, message losses and message delays-- to derive optimal strategies, bounds on the information requirements, and scaling properties of these bounds. Moreover, we present a specific network communication protocol which has the required scalable information characteristics and conclude with the results of experimental studies.

The Kansei testbed at the Ohio State University is designed to facilitate research on networked sensing applications at scale. Kansei embodies a unique combination of characteristics as a result of its design focus on sensing and scaling: (i) Heterogeneous hardware infrastructure with dedicated node resources for local computation, storage, data exfiltration and back-channel communication, to support complex experimentation, (ii) Time accurate hybrid simulation engine for simulating substantially larger arrays using testbed hardware resources, (iii) High fidelity sensor data generation and real-time data and event injection, (iv) Software components and associated job control language to support complex multi-tier experiments utilizing real hardware resources and data generation and simulation engines. In this paper, we present the elements of Kansei testbed architecture, including its hardware and software platforms as well as its hybrid simulation and sensor data generation engines.

We present a fast, local clustering service, FLOC, that partitions a multihop wireless network into nonoverlapping and approximately equal-sized clusters. Each cluster has a clusterhead such that all nodes within unit distance and some nodes within distance m of the clusterhead belong to the cluster. We show that, by asserting a stretch factor m ges 2, FLOC achieves locality of clustering and fault-local self-stabilization: the effects of cluster formation and faults/changes at any part of the network are contained within at most m + 1 units. Through simulations and experiments with actual deployments, we analyze the trade-offs between clustering time and the quality of clustering and suggest suitable parameters for FLOC to achieve a fast completion time without compromising the quality of the resulting clustering

Recent experiments have taken steps towards realizing the vision of extremely large wireless sensor networks, the largest of these being ExScal, in which we deployed about 1200 nodes over a 1.3 km by 300 m open area. Such experiments remain especially challenging because of: (a) prior observations of failure of sensor network protocols to scale, due to network faults and their spatial and temporal variability, (b) complexity of protocol interaction, (c) lack of sufficient data about faults and variability, even at smaller scales, and (d) current inadequacy of simulation and analytical tools to predict sensor network protocol behavior. In this paper, we present detailed data about faults, both anticipated and unanticipated, in ExScal. We also evaluate the impact of these faults on ExScal as well as the design principles that enabled it to satisfy its application requirements despite these faults. We describe the important lessons learnt from the ExScal experiment and suggest services and tools as a further aid to future large scale network deployments.

Project ExScal (for extreme scale) fielded a 1000+ node wireless sensor network and a 200+ node peer-to-peer ad hoc network of 802.11 devices in a 13km by 300m remote area in Florida, USA during December 2004. In comparison with previous deployments, the ExScal application is relatively complex and its networks are the largest ones of either type fielded to date. In this paper, we overview the key requirements of ExScal, the corresponding design of the hardware/software platform and application, and some results of our experiments.

The influence field of an object, a commonly exploited feature in science and engineering applications, is the region where the object is detectable by a given sensing modality. Being spatially distributed, this feature allows us to tradeoff nodal computation with network communication. By the same token, not only is its calculation subject to nodal failures and false detections, but also to channel fading and channel contention. In this paper, we study how to accurately and efficiently estimate the influence fields of objects in such an unreliable setting and how this reliable estimation of influence fields can be used to classify and track different types of objects. We derive, for node and network fault models, the necessary nodal density for reliably estimating the influence fields so that objects can be classified and tracked. We present four algorithmic techniques: temporal aggregation, probabilistic reporting, temporal segregation and spatial reconstruction, to deal with cases where the effective network density differs from this minimum. We provide corroboration of our analysis through field experiments with Mica2 sensor nodes wherever appropriate. Finally, we demonstrate how these results and techniques were applied to achieve reliable and efficient classification and tracking in a fielded system of 90 Mica2 sensor nodes that we called "A Line In The Sand'.

We present a fast local clustering service, FLOC, that partitions a multi-hop wireless network into nonoverlapping and approximately equal-sited clusters. Each cluster has a clusterhead such that all nodes within unit distance of the clusterhead belong to the cluster but no node beyond distance m from the clusterhead belongs to the cluster. By asserting m /spl ges/ 2, FLOC achieves locality: effects of cluster formation and faults/changes at any part of the network are contained within most m units. By taking unit distance to be the reliable communication radius and m to be the maximum communication radius, FLOC exploits the double-band nature of wireless radio-model and achieves clustering in constant time regardless of the network size. Through simulations and experiments with actual deployments, we analyze the tradeoffs between clustering time and the quality of clustering, and suggest suitable parameters for FLOC to achieve a fast completion time without compromising the quality of the resulting clustering.

IP over optical is being envisioned as one of the most attractive architectures for the new Internet. There have been various proposals in IETF and other international standards organizations regarding the interaction of IP routers and optical core networks. This paper describes the architectural alternatives for the integration of IP and DWDM networks using Multiprotocol Lambda Switching (MPLambdaS). The paper also addresses the issues of routing, signaling, control and survivability in an all-optical network.