Robot Navigation in Human Populated Unknown Environments based on Visual-Laser Sensor Fusion

Abstract

Robotic systems are progressively integrated in many parts of our lives to a point where their acceptance is now extensive inside our society. Nevertheless, their social skills are still very limited when they navigate in human populated environments. It is well established, that the robotic motion in any populated environment requires awareness in order for the robot to be socially compliant and not constitute any danger to its surroundings. For this purpose, we have developed a framework for robotic human-aware navigation that can also be used as an installation framework in a totally unknown environment. Our contribution also involves a 3d human pose estimation and a human laser tracking module that are based on the rgb-d and laser sensors respectively. The outcome of our framework results in the adjustment of the robot’s path in accordance to the social norms. This way, the robot can navigate in its environment without intruding the personal space of the people inside it or disquiet them in any way. The potential of our framework was investigated through experiments that were conducted both in a simulation and real environments. The real environment where we experimented on consists a warehouse, but the algorithm results indicate that it could be easily implemented in domestic environments to aid other tasks like assisting living.