Dynamic and Adaptive Sensor Operations Under Uncertainty
Abstract
Computational Methods for Decision Making (ONRBAA 15-001) Project Summary: Dynamic and Adaptive Sensor Operations Under Uncertainty J. Cole Smith (PI) The proposed research studies the deployment and operational strategy of sensors in un- certain environments. (The project is intended to redistribute funds for an existing project to the PI s new institution.) Our research examines sensors that are imperfect, and whose ability to detect targets degrades as a function of distance from the target. The problems considered are particularly relevant in the context of underwater sensing and communication, noting the speci c challenges in this domain pertaining to highly-limited bandwidth, power constraints, signal attenuation, and so on. The problems that we consider attempt to optimize one of two objectives. In the xed target objective, there exists a given set of target areas that are to be monitored by the deployed sensors. A target is successfully monitored if it is detected by at least one sensor. The goal is to maximize the expected number of targets (perhaps weighted by relative importances of the targets) that are monitored by the sensor network. In the adversarial evader objective, a mobile entity seeks to avoid detection by the sensors. With given knowledge of the sensor deployments, the adversary seeks an origin-destination path that maximizes its probability of detection. The objective is thus to deploy sensors in a manner that minimizes the adversary s maximum evasion probability. Distinct from previous work in this eld, we assume that sensors will occasionally fail, and that the network will need to dynamically recon gure itself in response. This recon guration requires several adjustments, including repositioning of sensors and rerouting communication patterns within the network. Moreover, this recon guration must be done in a distributed (decentralized) manner, without overloading sensor power limitations, and without exceeding the low bandwidths available in inter-sensor communications. Moreover, the sensor networks that we examine here may involve patrolling sensors whose movements are synchronized to more e ectively monitor a larger area using limited sensor resources. Recon guring such networks entails a signi cant challenge, which we propose to address in this project. Some of the primary challenges that will be undertaken with this research are (a) prescrib- ing e ective initial sensor deployments (including patrolling patterns where appropriate), (b) designing e cient algorithms for decentralized network recon guration, using distributed op- timization concepts that require little communication among sensors, and (c) linking these strategic and tactical decisions so that the initial deployment anticipates possible failures. The third task is vital, as it ensures that deployed sensor networks remain robust with respect to sensor failures. Our approach to this integration step will examine and augment state-of- the-art stochastic programming methods, and will permit us to link our deployment model with alternative sensor recon guration models developed by our colleagues (in addition to those that we develop as a part of the proposed research). Hence, the near-term importance of our research provides decision-making support for deployment and operations of under- water sensors in uncertain environments, and the long-term importance is that the proposed methods are adaptable to future concepts of operations involving autonomous agents.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Jun 03, 2016
- Source ID
- N000141612110
Entities
People
- J. Cole Smith
Organizations
- Clemson University
- Office of Naval Research
- United States Navy