A Dynamic Sensor Selection and Allocation Approach for Markov Localization

Markov localization has been proven to be an effective and efficient technique for determining the physical locations of an autonomous agent whose behavior is nondeterministic and perceptions of its environment are limited. In this paper, we apply the Markov localization approach to the problem of localizing multiple moving targets using distributed sensors in an environment where the targets move nondeterministically and sensory information is unreliable. For a single target, we propose heuristics to select appropriate sensors for tracking the target so as to optimize the tracking accuracy. For multiple targets, we propose heuristics to allocate the available sensors to individual targets so as to optimize the overall system performance. The heuristics are adaptive in that they select and allocate sensors based on the dynamic estimations of the target´s locations. We empirically showed these heuristics outperformed the random selection and allocation approaches