Self-Supervised Traversability Prediction by Learning to Reconstruct Safe Terrain

TL;DR

This paper introduces a self-supervised method for off-road terrain traversability prediction that learns from vehicle experience without human annotations, using autoencoders to distinguish safe from risky terrain based on reconstruction errors.

Contribution

The authors propose a novel self-supervised approach that predicts terrain risk levels by reconstructing masked vehicle trajectory regions, eliminating the need for manual annotations.

Findings

Achieved 81.1% and 85.1% accuracy in separating low- and high-risk terrains.

Effective across different models, bottleneck sizes, and testing sites.

Demonstrated robustness in off-road autonomous navigation scenarios.

Abstract

Navigating off-road with a fast autonomous vehicle depends on a robust perception system that differentiates traversable from non-traversable terrain. Typically, this depends on a semantic understanding which is based on supervised learning from images annotated by a human expert. This requires a significant investment in human time, assumes correct expert classification, and small details can lead to misclassification. To address these challenges, we propose a method for predicting high- and low-risk terrains from only past vehicle experience in a self-supervised fashion. First, we develop a tool that projects the vehicle trajectory into the front camera image. Second, occlusions in the 3D representation of the terrain are filtered out. Third, an autoencoder trained on masked vehicle trajectory regions identifies low- and high-risk terrains based on the reconstruction error. We evaluated our approach with two models and different bottleneck sizes with two different training and testing sites with a fourwheeled off-road vehicle. Comparison with two independent test sets of semantic labels from similar terrain as training sites demonstrates the ability to separate the ground as low-risk and the vegetation as high-risk with 81.1% and 85.1% accuracy.