Decentralised multi-sensor multi-target registration
Eric W. Nettleton · 2005
This paper presents theory and results of a gridlock implementation. The approach presented relies on tracking a small number of targets (position and velocity) and, simultaneously, all relevant sensor errors in the system. As it is formulated as an asynchronous solution, it lends itself to a decentralised implementation. The current system only tracks sensor offset and pointing errors, although extension to track range and range rate errors in radar-like sensors is straightforward. It was found that the performance of this type of registration technique depends on the number of targets (mutually) tracked, the type of target behaviour observed, target geometry and the time targets are in track. It also depends on the properties of the sensors used: more accurate sensors have more demanding requirements for registration, if their full benefit is to be realised. Predictable targets, tracked by all sensors are most likely to result in good convergence of the registration solution. (4 pages)