Rough Terrain Navigation for Legged Robots using Reachability Planning and Template Learning

Navigation planning for legged robots has distinct challenges compared to wheeled and tracked systems due to the ability to lift legs off the ground and step over obstacles. While most navigation planners assume a fixed traversability value for a single terrain patch, we overcome this limitation by proposing a reachability-based navigation planner for legged robots. We approximate the robot morphology by a set of reachability and body volumes, assuming that the reachability volumes need to always be in contact with the environment, while the body should be contact-free. We train a convolutional neural network to predict foothold scores which are used to restrict geometries which are considered suitable to step on. Using this representation, we propose a navigation planner based on probabilistic roadmaps. Through validation of only low-cost graph edges during graph expansion and an adaptive sampling scheme based on roadmap node density, we achieve real-time performance with fast update rates even in cluttered and narrow environments. We thoroughly validate the proposed navigation planner in simulation and demonstrate its performance in real-world experiments on the quadruped ANYmal.

I. Introduction

Navigation planning for legged robots has distinct challenges which are not present for other types of robots. While flying robots attempt to avoid any contact with the environment, ground robots by definition require contact with the ground to locomote. Compared to other types of ground robots, which have a constant contact patch with the ground, legged robots can overcome obstacles by lifting their legs. Most traditional navigation planning approaches assume a single traversability value for any given terrain patch, which they check against the footprint of the robot [1], [2]. These approaches are limiting for legged robots due to their ability to change their footprint and choose contact locations with the environment deliberately. Therefore, we have chosen to apply a different, simplified robot representation when planning for legged systems based on limb reachability abstractions [3]. We represent a robot as one collision volume for its torso, and one reachability volume for each of its limbs. When checking the feasability of a given robot pose, we expect the torso volume to be collision-free, while we enforce collision for the reachability volume, to ensure that the robot is able to make environment-contact with its legs.