WayFAST: Navigation with Predictive Traversability in the Field

1 Field Robotics Engineering and Science Hub (FRESH), Illinois Autonomous Farm, University of Illinois at Urbana-Champaign (UIUC), IL
2 EarthSense Inc., Champaign, IL, USA
3 DEVCOM Army Research Lab, Adelphi, MD, USA
4 Dept. of Aerospace Engineering, UIUC, IL, USA
Correspondence to {mvalve2, girishc}@illinois.edu
Published in the IEEE Robotics and Automation Letters with presentation in IROS 2022

We present a self-supervised approach for learning to predict traversable paths for wheeled mobile robots that require good traction to navigate. Our algorithm, termed WayFAST (Waypoint Free Autonomous Systems for Traversability), uses RGB and depth data, along with navigation experience, to autonomously generate traversable paths in outdoor unstructured environments. Our key inspiration is that traction can be estimated for rolling robots using kinodynamic models. Using traction estimates provided by an online receding horizon estimator, we are able to train a traversability prediction neural network in a self-supervised manner, without requiring heuristics utilized by previous methods. We demonstrate the effectiveness of WayFAST through extensive field testing in varying environments, ranging from sandy dry beaches to forest canopies and snow covered grass fields. Our results clearly demonstrate that WayFAST can learn to avoid geometric obstacles as well as untraversable terrain, such as snow, which would be difficult to avoid with sensors that provide only geometric data, such as LiDAR. Furthermore, we show that our training pipeline based on online traction estimates is more data efficient than other heuristic-based methods.

We tested WayFAST in a variety of environments

WayFAST Overview

WayFAST is a modular architecture. In our method, images are used to predict a local traversability map, which is used to generate a cost function for the model predictive control (MPC) block. MPC generates locally optimal goal-oriented trajectories of good traction that avoid obstacles.

To drive the robot on a safe path to reach the goal location, we formulate a nonlinear model predictive controller (NMPC) that leverages convolutional network traversability prediction. A kinodynamic model is used to predict the robot's states while network output is used to predict terrain traversability.