An Intelligent Vision Localization System of a Service Robot Nao

With the development of robot technology, home service robots are stepping into families. Localization and navigation are the important issues of the mobile robot research. In this paper, an intelligent vision localization system for obstacle avoidance and grasp task of an indoor service robot is developed. The system comprises of a ceiling camera which is placed overhead of the robot and the camera of a humanoid robot Nao. The ceiling camera provides global vision and the robot's camera gives local vision. In this research, global vision and local vision are integrated to improve the vision localization effect. To test the proposed system, we designed an experiment, in which robot Nao moves to a place, avoids collisions and performs the grasp task. The experiment consists of two parts: intelligent image space built by the ceiling camera and robot's image space built by the robot's own camera. In the intelligent image space, the system completes the rough visual location. And in the robot's image space, the robot accomplishes the precise positioning and adjusts itself to a more accurate pose for grasp task. In the second part, a simplified monocular vision positioning method is proposed and it offers good performance on accuracy. The experimental results show that the mean distance deviation of localization is 0.028 m. The system has been successfully tested for real time visual localization and grasp task.