VENRiR: Vision Enhance for Navigating 4-legged Robot in Rough Terrain

摘要

Nowadays, legged robots have become prevalent due to its versatility and ability to navigate in hard-to-access environments such as urban search and rescue mission or mudding area for agricultural purposes .Normally, quadruped robot's configuration is determined by force sensor attached at the leg position. However, in this paper, we present a framework for quadruped robot control by utilizing a visual system as exteroceptor to estimate robot's configuration and motor's torque sensors as proprioceptor for motion controller. Then, given the goal position, our designed motion controller will determine the best path and schedule suitable gaiting strategy such that legs' motion can compensate robot's body to level with the horizontal plane . As a result, it can compliantly balance its trunk while traversing rough terrain on the way to reach the destination. Two extensive experimental results: robot position control and orientation stabilization, conclude that our proposed control scheme performs effectively except for the sharp-turn corner. This i s due to the basic structure of our robot that does not have shoulder joint.