Real-time Autonomous Robot Navigation System with Collision Avoidance for NAO Robot

This thesis describes the development and performance of a full-stack autonomous navigation system. It integrates image and depth-based environmental perception, path planning, and walking control of the NAO humanoid robot for navigation in unknown terrains. The perception module detects walkable surfaces from the camera image and time-to-collide with obstacles from the point cloud data. The path to move from the current position to the destination is solved using a search algorithm. The robot moves along the defined path by the walking controller module. This navigation system is tested on the NAO robot in simulation, and performance results are presented.