Learning Smooth State-Dependent Traversability from Dense Point Clouds

2025-06-04

AI Summary

In off-road autonomous driving, terrain traversability is highly dependent on vehicle approach direction; however, existing methods treat orientation as a discrete input, requiring extensive labeled data and incurring high inference overhead. This paper proposes the first dense point-cloud-based framework explicitly modeling orientation-sensitive traversability. We design a geometry-aware network that, for the first time, employs Fourier bases on the 1-sphere to explicitly represent smooth, orientation-dependent traversability risk distributions—enabling single-pass inference, continuous angular queries at arbitrary orientations, and strong generalization. In simulation, our method achieves 91% success rate navigating a 40-meter rocky terrain—outperforming baselines by 18 percentage points. Hardware experiments further validate robustness in real-world scenarios. The core contribution lies in elevating orientation modeling from discrete inputs to a continuous analytical function, thereby jointly optimizing accuracy, computational efficiency, and generalizability.

Technology Category

• Search and Optimization: Mixed Discrete/Continuous Search
• Intelligent Robots: Localization, Mapping, and Navigation
• Planning, Routing, and Scheduling: Model-Based Reasoning

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web
• Systems and Infrastructure for Web, Mobile and WoT: Experiences and lessons learnt from Web-based algorithms and system deployments

Abstract

A key open challenge in off-road autonomy is that the traversability of terrain often depends on the vehicle's state. In particular, some obstacles are only traversable from some orientations. However, learning this interaction by encoding the angle of approach as a model input demands a large and diverse training dataset and is computationally inefficient during planning due to repeated model inference. To address these challenges, we present SPARTA, a method for estimating approach angle conditioned traversability from point clouds. Specifically, we impose geometric structure into our network by outputting a smooth analytical function over the 1-Sphere that predicts risk distribution for any angle of approach with minimal overhead and can be reused for subsequent queries. The function is composed of Fourier basis functions, which has important advantages for generalization due to their periodic nature and smoothness. We demonstrate SPARTA both in a high-fidelity simulation platform, where our model achieves a 91% success rate crossing a 40m boulder field (compared to 73% for the baseline), and on hardware, illustrating the generalization ability of the model to real-world settings.

Problem

• Estimating traversability based on vehicle approach angle
• Reducing computational inefficiency in planning via smooth analytical functions
• Generalizing model performance to real-world off-road autonomy scenarios

Innovation

• Estimates traversability from point clouds
• Uses Fourier basis for smooth risk prediction
• Achieves high success in simulation and hardware