Anytime Resource Optimization with Sliding Performance for Mission Planning and Sensor Development
Abstract
Statement of Work: The proposed research will develop optimization tools for mission planning and sensor management. Objective: The proposed research is based on a need for resource optimization tools for sensor management (e.g., a fleet of UXVs) whose performance slides gradually as a function of the complexity of the problem and the communication connectivity between resources. Furthermore, these tools must be anytime, that is, the quality of the solutions they return must also depend on the time available for computation. In some situations the decisions must made within seconds, and in others there are hours available for optimization. Anytime algorithms are algorithms that generate a first, possibly highly suboptimal, solution quickly, and then use the remaining time to improve it until either a provably optimal solution is achieved or time available for computation runs out. Approach: The approach of the project can be broken down into the following key steps: Anytime approaches to asset allocation and mission planning with sliding performance with respect to problem complexity and size of the fleet: We will develop an algorithmic framework which is anytime - generates the best solution it can find within whatever time it is given for optimization and improves the solution further during execution - and whose performance degrades gracefully as the problem complexity increases and/or size of the fleet increases. Approaches to asset allocation and mission planning with sliding performance with respect to to network connectivity: We will develop approaches that utilize whatever and whenever connectivity is available and whose performance degrades gracefully as the quality of the network connectivity degrades. The quality of the network connectivity is characterized by the availability of communication, available bandwidth, reliability and latency. Probabilistic framework for cooperative search, identication, location and tracking of targets: We will develop the mathematical framework and the algorithms for search, localization and tracking for an unknown number of targets within a known map using a team of heterogeneous vehicles equipped with noisy, limited field-of-view sensors. This framework will employ the resource optimization tools we develop as part of the above-stated objectives. Scalability analysis of the developed algorithms and methodology with respect to the size of the fleet: We will study, both analytically and experimentally in simulation, the scalability of our approaches with respect to the size of the team. Experimental study and validation: We will study the performance of our approaches through extensive evaluation using Hardware-in the-loop simulation setup that we have at the GRASP laboratory, University of Pennsylvania. We will also validate our approaches on a team of physical fixed-wing aircrafts using Rapid Flight Test Prototyping System (RFTPS) developed at Naval Postgraduate School by I. Kaminer whose is a collaborator on this proposal. Overall Merit and ONR Mission/Relevance: The Navy is moving towards increasing use of unmanned sensor platforms. The proposed research is aimed at developing state-of-the-art tools for mission planning and management of these assets.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Aug 08, 2016
- Source ID
- N000141512129
Entities
People
- Maxim Likhachev
Organizations
- Massachusetts Institute of Technology
- Office of Naval Research
- United States Navy