Performance Evaluation of ROS Local Trajectory Planning Algorithms to Social Navigation

摘要

Accuracy and safety are necessary characteristics in social navigation and still constitute a challenge. The ROS Navigation Stack (RNS) allows the variation of local path planning methods through plugins for navigation. This paper brings you the comparison of those methods, which are directly connected with the safety and naturalness of the robot. Therefore, four different methods were compared by varying the sensors and the simulated environment. A thousand experiments were performed for each combination using the standard parameters of each method in a total of 24000 experiments. This paper concluded that the Elastic Band (EBand) method presents more safety and accuracy than the Dynamic window approach (DWA), method commonly used in several robots that participate in RoboCup@home, so it is more suitable for social navigation - reaching 90% accuracy in some cases and collision rate below 5%.