Self-adaptively pruning and renovating the hypothesis tree of dynamic world

This paper presents a new self-adaptive pruning and renovation strategy for modeling dynamic world, which is a crucial problem for mobile robots to build and maintain a map of its environment by means of a Bayesian multiple hypothesis approach. This strategy mainly consists of two parts: pruning and renovation. Pruning eliminates those branches of the hypothesis tree with lower probabilities. Once the hypothesis branch with the greatest probability is not identical with the actual measurements, other hypotheses will be compared with the actual measurements in turn due to their probabilities. If criteria are satisfied, the strategy will retain it and prune all other branches.