Dynamic and Adaptive Sensor Operations Under Uncertainty
Abstract
Statement of Work:The PI will develop optimization-based tools to aid in the deployment and operational strategy of sensors in uncertain and changing environments.Objective:The PI proposes to enhance the study of sensor deployment and operation problems to encompass those that are of practical interest to the defense community, and are of particular relevance in UUV deployment operations. In particular, he proposes to examine the following facets of modern sensor deployment and operations problems.1. The sensors may be deployed either with the objective of monitoring the greatest number of targets possible, or with the objective of minimizing an adversary s maximum probability of evasion. Both problems become more difficult when the functionality of the sensors degrades as a function of its distance to the target, and the latter problem has, in particular, not received ample attention in the literature relative to its real-world importance.2. The sensors themselves are assumed to be mobile, in order to dynamically accommodate for sensor failures. We will examine the case in which these mobile sensors are stationary while performing monitoring operations (and whose mobility is used only to reposition themselves when other sensors fail), and the case in which sensors patrol certain routes while monitoring.3. Uncertainty can manifest itself in many ways, from the random failure of sensors, to erroneous messages caused by noisy acoustic signals, to interference in sensing operations. The proposed studies will examine deployment and operations models that are robust with respect to these uncertainties.Approach:The approach will be focused on three specific challenges related to limited sensor communication.(a) The PI will develop various simple protocols for sensor reconfiguration, based on geometric and duality principles, and will compare their effectiveness with a centralized control scheme. (b) Given a decentralized repositioning mechanism, the goal is to optimize the deployment of sensors underuncertainty, to maximize the expected efficiency of the sensor network. The PI proposes to develop core methods that are flexible enough to accommodate very complex, "black-box" style protocols that govern decentralized sensor control. Hence, the research lends itself well to integration with complementary research projects funded by the ONR.(c) Communication among sensors is important not only to relay observed information to a base station, but also to alert each other to changing environmental conditions and to failed sensors. Message routing among sensorscontributes to battery drain and in many settings (e.g., when sensor batteries cannot be replaced) may lead to the premature failure of a sensor. The PI will thus embed the challenges of energy-conserving routing mechanisms that guard against overloading some sensors with excessive communication load. In addition to the practical importance of our proposed studies, the research proposes to advance the state-of-the-art in spatial interdiction models and stochastic programming methodology.Overall Merit and ONR Mission/Relevance:The proposed work is innovative in that it develops new techniques for solving a difficult class of non-convex stochastic optimization problems. The core problem of sensor deployment and autonomous operation is of direct concern in the Navy. Moreover, the proposed research focuses on developing methods for both strategic and tactical operations, and develops the mathematical and algorithmic foundations required to link them.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Sep 23, 2016
- Source ID
- N000141612802
Entities
People
- Guanghui Lan
Organizations
- Georgia Tech Research Corporation
- Office of Naval Research
- United States Navy