Learning When to Jump for Off-road Navigation

TL;DR

The paper tackles the challenge of off-road navigation where terrain safety and efficiency depend on the robot's motion, not just perception. It introduces Motion-Aware Traversability (MAT), a velocity-dependent per-pixel Gaussian representation $T(\mathbf{x},v)=A(\mathbf{x}) \exp(-\frac{(v-\mu(\mathbf{x}))^2}{2\,\sigma(\mathbf{x})^2})$ learned from self-supervised vehicle-terrain interactions. MAT enables real-time, motion-conditioned planning by providing a closed-form terrain cost $T(\mathbf{x},v)$ that can be efficiently evaluated for any velocity during MPPI optimization. The approach demonstrates substantial reductions in detour distance (up to 75%), improved traversal time and energy, and robust performance in simulation and real-world ditch and obstacle scenarios, highlighting its practical impact for agile and safe off-road navigation. Future work aims to extend beyond Gaussian modeling and incorporate additional motion factors to further enhance robustness and expressiveness.

Abstract

Low speed does not always guarantee safety in off-road driving. For instance, crossing a ditch may be risky at a low speed due to the risk of getting stuck, yet safe at a higher speed with a controlled, accelerated jump. Achieving such behavior requires path planning that explicitly models complex motion dynamics, whereas existing methods often neglect this aspect and plan solely based on positions or a fixed velocity. To address this gap, we introduce Motion-aware Traversability (MAT) representation to explicitly model terrain cost conditioned on actual robot motion. Instead of assigning a single scalar score for traversability, MAT models each terrain region as a Gaussian function of velocity. During online planning, we decompose the terrain cost computation into two stages: (1) predict terrain-dependent Gaussian parameters from perception in a single forward pass, (2) efficiently update terrain costs for new velocities inferred from current dynamics by evaluating these functions without repeated inference. We develop a system that integrates MAT to enable agile off-road navigation and evaluate it in both simulated and real-world environments with various obstacles. Results show that MAT achieves real-time efficiency and enhances the performance of off-road navigation, reducing path detours by 75% while maintaining safety across challenging terrains.

Figures (7)

• Figure 1: Challenging terrains often require motion strategies that span from aggressive maneuvers to cautious crawling. Our Motion-Aware Traversability (MAT) model enables the vehicle to decide when to accelerate to jump over a ditch, slow down to crawl over curbs, and steer to avoid obstacles. In contrast, baseline methods assign uniformly high costs to features like ditches, causing the vehicle to simply avoid them rather than selecting motion strategies that could traverse them more efficiently.
• Figure 2: The Architecture of the proposed pipeline. Given a local perception input (a LiDAR-based height map in practice), the neural network performs one-shot inference to generate the Motion-aware Traversability (MAT) map, where each cell encodes a Gaussian function of velocity. The local planner (e.g., MPPI) then samples trajectories based on vehicle dynamics, continuously querying the MAT map for terrain costs with position-velocity pairs to optimize motion in real time.
• Figure 3: Navigation system diagram showing data flow from LiDAR sensing to actuator commands, with ROS nodes performing perception, MAT estimation, planning, and control.
• Figure 4: Qualitative comparison between MAT and the PO-Trav in the obstacle traversal task. For each method, the vehicle trajectories with annotated velocities are shown for three obstacle types and three operating speeds. The velocity at which the obstacle first enters perception range is highlighted as a reference, and the full velocity profile is plotted for each test.
• Figure 5: Simulation maps and qualitative results for the long-range navigation task. The top row shows three modified maps with introduced obstacles blocking the routes. The bottom row presents vehicle trajectories under MAT and the PO-Trav that exhibit distinct behavioral differences, with the corresponding obstacles highlighted in bold on the maps.