Using Bacteria Foraging Optimization in Gesture Reconfiguration from Joint Failure for Semi-Humanoid Robot

Abstract

Joint failure problem in humanoid robots could often occur in real application. An approach for fault recovery is performing gesture reconfiguration on a redundant manipulator of humanoid robots. In this study, we proposed a gesture reconfiguration method based on bacteria foraging optimization algorithm. After joint failure, we locked the failed joint in place and used the BFOA-based gesture reconfiguration to find a new suitable gesture configuration for the remaining joints. We demonstrated the gesture reconfiguration from joint failure on the Namo robot, which is a semi-humanoid robot for performing emblematic gestures as a receptionist robot. Four emblematic gestures, including the Thai greeting, salute, bye, and side invite gestures, were demonstrated. The experimental results showed that the BFOA-based gesture reconfiguration could generate a suitable gesture configuration for all four gestures; the resulting gestures are close to the original gestures. We investigated further on tuning the number of population size in BFOA. We varied the number of population size as 15, 25, 50, and 100 and found that although the larger population size could result in better gestures than the smaller size, the improvement in gesture similarity were not significant so it may be more efficient to use the smaller size. This could be owing to the performance of bacteria foraging optimization algorithm.