Centralized and Decentralized Kalman Filter Techniques for Tracking, Navigation, and Control. Revision
Abstract
A review of some estimation basics is followed by illustrative applications of Kalman filters for stationary and maneuvering targets. The variable dimension of Kalman filter is used for the maneuvering target. The performance of the nearest neighbor standard filter is compared to that of the probabilistic data association filter for tracking a target in clutter. Multi-target tracking, using sonar sensors to estimate an autonomous robot's distance from walls, is applied to the navigation problem. The Kalman filter equations can be completely decentralized and distributed among the nodes of a multi-sensor system. Each sensing node implements its own local Kalman filter, arrives at a partial decision, and broadcasts it to every other node. Each node then assimilates this received information to arrive at its own local but optimal estimate of the system state. An appendix contains brief implementational notes. (rrh)
Document Details
- Document Type
- Technical Report
- Publication Date
- May 01, 1989
- Accession Number
- ADA214245
Entities
People
- Barry Steer
- Bobby Rao
- Chris Brown
- Hugh Durrant-whyte
- John J. Leonard
Organizations
- University of Rochester