Data association for a hybrid metric map representation

Abstract

This paper presents an approach to solve the data association problem for a hybrid metric map representation. The hybrid metric map representation uses Voronoi diagram to partition the global map space into a series of local subregions, and then a local dense map is built in each subregion. Finally the global feature map and the local maps make up of the hybrid metric map, which can represent all the observed environment. In the proposed map representation, there exists an important property that global feature map and local maps have clear one-to-one correspondence. Benefited from this property, an identifying rule of the data association based on compatibility testing is proposed. The identifying rule can efficiently reject the wrong data association hypothesis in the application of dense environment. Two experiments validated the efficiency of data association approach and also demonstrated the feasibility of the hybrid metric map presentation.