The goal of our project is to develop a solution that allows a quadruped robot to avoid footholds that may cause it to fall over. This involves:
Developing a controller that allows the quadruped to balance on 3 legs
Estimating the stability of a foothold by probing it with force sensors
Walking the robot using MPC controllers for quadruped locomotion
Quadrupeds can be used for real-world applications such as search and rescue missions, and construction site inspection. These tasks often involve environments that contain unstable terrain. For instance, a quadruped might be deployed into a burning building to search for survivors, and must detect whether stepping in a certain location would cause the robot to fall over. In such situations, vision and depth se nsor systems may be able to classify the type of terrain, but these sensor systems have no way of determining if the terrain identified has the structural strength required to support the quadruped. Avoiding unstable footholds in these mission-critical situations is especially important. Our project seeks to develop a solution that allows a quadruped to probe if a foothold is stable. If the foothold is deemed stable, the robot will use it, and if it is not stable, the robot will avoid it. Additionally, during probing, the robot will need to be able to balance on three legs, which will be solved via the quadruped's center of mass calculations.