ACTIVE AND STABLE SLAM BASED ON MULTI-OBJECTIVE OPTIMIZATION

In the EKF-based simultaneous localization and mapping (SLAM), unstable observations are likely to be generated due to the linearization operation of EKF, which will increase the errors of localization and mapping. In this paper, a novel active SLAM algorithm considering observation stability is proposed. We analyse the impacts of observation distance on the estimation process of EKF. Then, the active SLAM is converted into a problem of multi-objective optimization, in which the objective function includes not only the accuracy of SLAM and time information gain, but also the observation distance. The robot chooses optimal control inputs to decrease uncertainty, acquire more environmental information, and obtain more stable observations simultaneously. In such a way, accurate and stable map building can be achieved. Simulations and experimental results demonstrate that the performance of the proposed algorithm is superior to that of other methods.