Real-Time Visual Perception System for a Robotic Rat

摘要

The visual sensing ability of the robotic rat will be conducive to the biomimetic interaction between the robot and rats. However, most of the existing small-sized robotic rats do not have perception system, limiting their ability to perceive their surroundings or tracing the object. To address this problem, we developed a visual perception system for the robot, which included a control and drive system, a stereo camera integrated in the robot's head, and a wireless image transmission system. We also analyzed wheel odometry model and integrate it into the robot. The whole system can achieve real-time data acquisition and processing. In the experimental scene, we extracted the number of dynamic rats feature points and analyzed the impact on environment modeling, and these feature points were easy to interfere with the robot's perception and modeling of the environment. Additionally, our system was capable of SLAM (Simultaneous Localization and Mapping) when tested by using the RTAB-Map (Real-Time Appearance-Based Mapping) algorithm. The experimental results showed that the trajectory was basically coherent and finally returns to the origin with less drift.