Intelligent Navigation of Humanoids in Cluttered Environments Using Regression Analysis and Genetic Algorithm

In this study, two navigational controllers have been developed for the path planning of single as well as multiple humanoid robots in a cluttered environment using classical and computational intelligence approaches. Regression analysis and genetic algorithm have been used to design the proposed controllers. The regression controller is developed based on the left, right and front obstacle distances referenced from the humanoid’s current position and orientation and aims to calculate an optimized turning angle for minimum path length. The genetic algorithm controller is developed based on the nearest obstacle distance and goal position relative to the current position and orientation of the humanoid and aims to calculate its next best position for an optimized path length. To avoid inter-collision in the navigation of multiple humanoids, a Petri-Net controller has been implemented. The proposed algorithms have been successfully validated through multiple simulations in V-REP software. To test the effectiveness of the controllers, real-time experiments have also been conducted with NAO humanoids, and the results were compared with those obtained from the simulations. It was found that the experimental results closely resemble the simulations in terms of trajectories followed, path length covered and overall time taken with an acceptable error limit. The proposed navigational technique has also been compared with other existing navigational approaches to validate its effectiveness. Finally, it was concluded that the proposed navigational controllers are efficient in the path planning and obstacle avoidance and can be implemented on humanoid navigation in complex environments.