We study the problem of exploring an unknown environment using a single robot. The environment is large enough (and possibly dynamic) that constant motion by the robot is needed to cover the environment. We term this the dynamic coverage problem. We present an efficient minimalist algorithm which assumes that global information is not available to the robot (neither a map, nor GPS). Our algorithm uses markers which the robot drops off as signposts to aid exploration. We conjecture that our algorithm has a cover time better than O(n log n), where the n markers that are deployed form the vertices of a regular graph. We provide experimental evidence in support of this conjecture. We show empirically that the performance of our algorithm on graphs is similar to its performance in simulation.

Our research is focused around idea of using sensor network and mobile robots cooperatively for solving various tasks. Moreover, sensor networks and mobile robots both benefit from such collaboration. One of the examples of such "symbiotic" approach is solution to the mobile robot navigation problem. The task is to detect the area requiring robot"s presence in the environment and then to guide the robot into that area. We propose solution to such problem and present results from real-world experiments conducted at Intel research facilities in Hillsboro, Oregon.

In many sensing applications, including environmental monitoring, measurement systems must cover a large space with only limited sensing resources. One approach to achieve required sensing coverage is to use robots to convey sensors within this space. Planning the motion of these robots -- coordinating their paths in order to maximize the amount of information collected while placing bounds on their resources (e.g., path length or energy capacity) -- is a NP-hard problem. In this poster, we present an efficient path planning algorithm that coordinates multiple robots, each having a resource constraint, to maximize the “informativeness" of their visited locations. In particular, we use a Gaussian Process to model the underlying phenomenon, and use the mutual information between the visited locations and remainder of the space to characterize the amount of information collected. We provide strong theoretical approximation guarantees for our algorithm by exploiting the sub-modularity property of mutual information. We provide an empirical analysis of our algorithm from field experiments, using Networked Info Mechanical Systems (NIMS) family of robotic systems. The NIMS family of sensing systems, together with an efficient experimental design approach that involves phenomena modeling, enabled the first high resolution imaging of several important scientific phenomena such as contaminant concentration and algal bloom dynamics. This work is currently being applied to survey entire river systems in interdisciplinary investigations providing scientists with important new characterization of primary national water resources.

Background and purpose

Methods

Results

Conclusions

Many environmental applications require high temporal frequency (rapidly changing) and spatially distributed phenomena to be sampled with high fidelity. This requires mobile sensing elements to perform guided sampling in regions of high variability. We propose a multiscale approach for efficiently sampling such phenomena. This approach introduces a hierarchy of sensors according to the sampling fidelity, spatial coverage, and mobility characteristics. In this paper, we report the development of a two-tier multiscale system where information from a low-fidelity, high spatial (global) sensor actuates a mobile robotic node, carrying a high-fidelity, low spatial coverage (spot measurement) sensor, to perform guided sampling in the regions of high phenomenon variability. As a case study of the proposed multiscale paradigm, we investigated the spatiotemporal distribution of the light intensity in a forest understory. The performance of the multiscale approach is verified in simulation and on a physical system. Results suggest that our approach is adequate for the problem of high-frequency spatiotemporal phenomena sampling and significantly outperforms traditional sampling approaches such as a raster scan.

Recent advancement in microsensor technology permits miniaturization of conventional physiological sensors. Combined with low-power, energy-aware embedded systems and low power wireless interfaces, theses sensors now enable patient monitoring in home and workplace environments in addition to the clinic. Low energy operation is critical for meeting long operating lifetime requirement; an energy-aware wearable system is therefore particularly beneficial to adaptively profile and manage energy utilization. Furthermore, important challenges appear as some of these important physiological sensors, such as electrocardiographs (ECG), introduce large energy demand (because of the need for high sampling rate and resolution) and limitations (due to reduced convenience of user wearability). Energy usage of the distributed sensor systems may be reduced by activating and deactivating sensors according to real-time measurement demand as well as energy consumption characteristics. Our results show that with proper adaptive measurement scheduling, an ECG signal from a subject may be needed for analysis only at certain times, such as during or after an exercise activity. This demonstrates that autonomous systems may rely on low energy cost sensors combined with real time computation to determine patient context with high certainty diagnostics and apply this information to properly schedule use of high cost sensors (e.g. ECG sensor systems).

We have implemented a wearable system based on standard widely-used handheld computing hardware components. This system relies on a new software architecture and an embedded inference engine developed for theses standard platforms. The performance of the system is evaluated using experimental data sets acquired for subjects wearing this system during an exercise sequence. This same approach can be used in context-aware monitoring of diverse physiological signals in a patient’s daily life. Furthermore, a new energy-aware wearable system is introduced. It is capable of performing real-time energy profiling on major components through a convenient software interface. Exploring the techniques on how to utilize this energy information and optimize the existing context-aware algorithm is the focus of future work.

In many sensing applications, including environmental monitoring, measurement systems must cover a large space with only limited sensing resources. One approach to achieve required sensing coverage is to use robots to convey sensors within this space.Planning the motion of these robots – coordinating their paths in order to maximize the amount of information collected while placing bounds on their resources (e.g., path length or energy capacity) – is a NP-hard problem. In this paper, we present an efficient path planning algorithm that coordinates multiple robots, each having a resource constraint, to maximize the “informativeness” of their visited locations. In particular, we use a Gaussian Process to model the underlying phenomenon, and use the mutual information between the visited locations and remainder of the space to characterize the amount of information collected. We provide strong theoretical approximation guarantees for our algorithm by exploiting the submodularity property of mutual information. In addition, we improve the efficiency of our approach by extending the algorithm using branch and bound and a region-based decomposition of the space. We provide an extensive empirical analysis of our algorithm, comparing with existing heuristics on datasets from several real world sensing applications.

This paper reports the first application of iterative experimental design methodology for high spatiotemporal resolution characterization of river and lake aquatic systems performed using mobile robot sensing systems. Both applications involve dynamic phenomena spread over large spatial domain: 1) Characterization of contaminant concentration and flow at the confluence of two major rivers displaying dynamics due to flow of the water; and 2) Characterization of rapidly evolving biological processes such as phytoplankton dynamics in a lake system. We describe the development and application of a new general purpose method for mobile robot sensing in such environments - Iterative experiment Design for Environmental Applications (IDEA). IDEA introduces in-field adaptation of mobile robotic sensing system. Analysis of the complex spatial and temporal structures associated with each observed environment is presented. Detailed characterization of the observed environment using IDEA methodology is used as an informed prior to improve the performance of the existing adaptive experimental design approaches for mobile robotic systems - stratified adaptive sampling and hierarchical non-stationary Gaussian Processes.

Distributed, high-density spatiotemporal observations are proposed for answering many river related questions, including those pertaining to hydraulics and multi-dimensional river modeling, geomorphology, sediment transport and riparian habitat restoration. In spite of the recent advancements in technology, currently available systems have many constraints that preclude long term, remote, autonomous, high resolution monitoring in the real environment. We present here a case study of an autonomous, high resolution robotic spatial mapping of cross-sectional velocity and salt concentration in a river basin. The scientific objective of this investigation was to characterize the transport and mixing phenomena at the confluence of two distinctly different river streams - San Joaquin River and its tributary Merced River. Several experiments for analyzing the spatial and temporal trends at multiple cross-sections of the San Joaquin River were performed during the campaign from August 21-25, 2006. These include deterministic dense raster scans and in-field adapted experimental design. Preliminary analysis from these experiments illustrating the range of investigations is presented with the focus on adaptive experiments that enable sparse sampling to provide larger spatial coverage without discounting the dynamics in the phenomena. Lessons learned during the campaign are discussed to provide useful insights for similar robotic investigations in aquatic environments.

Distributed embedded sensor networks are now being successfully deployed in environmental monitoring of natural phenomena as well as for applications in commerce and physical security. Distributed architectures have been developed for cooperative detection, scalable data transport, and other capabilities and services. However, the complexity of environmental phenomena has introduced a new set of challenges related to sensing uncertainty associated with the unpredictable presence of obstacles to sensing that appear in the environment. These obstacles may dramatically reduce the effectiveness of distributed monitoring. Thus, a new distributed, embedded, computing attribute, self-awareness, must be developed and provided to distributed sensor systems. Selfawareness must provide the ability for a deployed system to autonomously detect and reduce its own sensing uncertainty. The physical constraints encountered by sensing require physical reconfiguration for detection and reduction of sensing uncertainty. Networked Infomechanical Systems (NIMS) consisting of distributed, embedded computing systems provides autonomous physical configuration through controlled mobility. The requirements that lead to NIMS, the implementation of NIMS technology, and its first applications are discussed here.