TOP-Nav: Legged Navigation Integrating Terrain, Obstacle and Proprioception Estimation

TL;DR

T TOP-Nav achieves open-world navigation that the robot can handle terrains or disturbances beyond the distribution of prior knowledge and overcomes constraints imposed by visual conditions.

Abstract

Legged navigation is typically examined within open-world, off-road, and challenging environments. In these scenarios, estimating external disturbances requires a complex synthesis of multi-modal information. This underlines a major limitation in existing works that primarily focus on avoiding obstacles. In this work, we propose TOP-Nav, a novel legged navigation framework that integrates a comprehensive path planner with Terrain awareness, Obstacle avoidance and close-loop Proprioception. TOP-Nav underscores the synergies between vision and proprioception in both path and motion planning. Within the path planner, we present and integrate a terrain estimator that enables the robot to select waypoints on terrains with higher traversability while effectively avoiding obstacles. In the motion planning level, we not only implement a locomotion controller to track the navigation commands, but also construct a proprioception advisor to provide motion evaluations for the path planner. Based on the close-loop motion feedback, we make online corrections for the vision-based terrain and obstacle estimations. Consequently, TOP-Nav achieves open-world navigation that the robot can handle terrains or disturbances beyond the distribution of prior knowledge and overcomes constraints imposed by visual conditions. Building upon extensive experiments conducted in both simulation and real-world environments, TOP-Nav demonstrates superior performance in open-world navigation compared to existing methods.

Figures (12)

• Figure 1: TOP-Nav achieves open-world navigation in both simulation and the real world. The robot plans an obstacle-free path on terrains with better traversability. The robot rapidly estimates its traversability for novel terrains based on proprioception experience.
• Figure 2: TOP-Nav framework: the path planner synthesizes ${\bm{M}_T, \bm{M}_O, \bm{M}_P, \bm{M}_G}$ into a combined cost map, from which it computes waypoints based on the overall cost considerations. The controller tracks the desired velocity and provide motion evaluations for the proprioception advisor.
• Figure 3: (a) The proposed terrain estimator incorporates a visual estimator and online corrections. (b) The motion evaluations respond rapidly when the robot encounters difficult terrains.
• Figure 4: Each navigation cell consists of randomly generated challenging terrains with distinct traverse difficulty, which is marked by the irregularity and complexity of the terrain. The proposed terrain awareness navigation framework plans an optimal path to navigate challenging terrains. The robot demonstrates the capability to recover from unexpected obstacles or irregular terrains with the proprioception advisor.
• Figure 5: (A) The terrain classifier does not include high-cost gravel in prior. (B) The robot encounters terrains with no prior knowledge, including a slippery detergent surface. We evaluate the effectiveness of the proposed terrain estimator within TOP-Nav in the experiments.