Terrain Classification for a Quadrupedal Robot Using Proprioceptive Sensing

Abstract

Quadrupedal robots can traverse a wider range of terrain types than their wheeled counterparts, but these robots do not perform the same on all terrain types. These robots are prone to undesirable behaviours like sinking and slipping on challenging terrain. To combat this issue, we can implement a terrain classifier and use the output to compute the best path for the robot to navigate. The work presented here is a terrain classifier developed for a Boston Dynamics Spot robot. The quadruped provides us with over 104 measured signals describing the position and speed of the robot and each of its four legs. The developed terrain classifier combines dimensionality reduction techniques to extract the relevant information from the signals and then applies a classification technique to differentiate terrain based on traversability. The resulting terrain classifier can identify three different terrain types with an accuracy of 93%. Ongoing work aims to get the classifier running in real time and generate cost maps to inform future path planning.