Dynamic Joint Nearest Neighbor Algorithm

摘要

'Nearest Neighbor'(NN) is a widely used data association algorithm in SLAM community.It owns the merit of low computational complexity while maintaining the demerit that its accuracy is susceptible to environments.Two improvements are introduced to enhance the robustness of NN to environments.One is clearing up the interference among all observation mates with a view to the relativity of all observations.The other is dynamically filtrating spurious features in observed features with the association results of measurements over multiple frames.What's more,data association is limited in potential local region,which is determined by pose of the robot and effective range of the sensor.Thus,computational efficiency of NN is improved greatly.Simulative and experimental results indicate that the presented 'Dynamic Joint Nearest Neighbor'(DJNN) performs well on both accuracy and computational complexity.And it is of wonderful value for real applications.