Data association in dynamic environments using a sliding window of temporal measurement frames

Correct data association is critical for the success of feature based simultaneous localization and mapping (SLAM) of autonomous vehicles or mobile robots. Incorrect associations result in map inconsistency and inaccurate path estimates. Numerous data association techniques proposed in the literature for SLAM assumes a static environment. Ignoring the effects of moving or dynamic objects leads to catastrophic failures. This work, proposes a new multiple frame batch temporal consistency criterion for data association in feature based SLAM in dynamic environments. Simulations and experimental results are presented to demonstrate the effectiveness of the algorithm.