For a quadruped robot, path planning presents a unique challenge. For an autonomous car, one needs to consider constraints like the size of the robot for avoiding obstacles and wheel size for turning radius. Compared to these systems, a quadruped contains more choices for a single action due to the high number of DOFs in the system. For example, a quadruped requires one to plan foot placement when moving from one location to another, which must be repeated for each leg, as well as assigning specific locations to four legs within areas of possible placement. Thus, it is easy to see that the dimensionality of the problem increases greatly along with the complexity of a local planner, thus requiring unique planning methods. This project seeks to implement a path planning algorithm for a quadruped robot on a simplified terrain model consisting of discrete allowed foot locations (“pegs”) with motion constrained by leg lengths and geometry. In this video, we discuss the implementation of a hierarchical planner for navigating this quadruped over terrain with both random and patterned valid foot placements.
