Research Analyzer
← Back ICRA 2026

How Bumpy Is It? Incremental Online Learning of Terrain-Induced Bumpiness Costs for Off-Road Vehicles

Haoyu Yuan, Tianwei Niu, Shengshan Ma, Runjiao Bao, Shoukun Wang

PDF

AI summary

Key figure (auto-extracted from paper)
The framework enables off-road vehicles to continuously learn and predict terrain bumpiness costs in real-time using only onboard IMU and sensor data, without manual annotations.
Off-road autonomous driving Terrain bumpiness Online learning IMU-based perception Incremental mapping Multimodal segmentation

Problem

Current off-road terrain perception relies on semantic segmentation that ignores physical drivability attributes, while supervised learning is hindered by costly annotations and an inability to generalize to unseen terrain types.

Approach

The system continuously learns terrain bumpiness costs online by deriving pseudo-labels from onboard IMU data, segmenting terrain with a lightweight multimodal model, and incrementally updating a cost map as new terrain types are encountered.

Key results

  • Real-time bumpiness cost prediction without pretraining
  • IMU-derived pseudo-labeling for cost computation
  • Incremental spatio-temporal updates for unseen terrain classification
  • Continuous cost mapping refinement during operation

Why it matters

It enables safe, real-time autonomous navigation for heavy off-road vehicles in unstructured environments by providing accurate traversability costs without costly data collection or retraining.

Abstract

Stable autonomous driving in unstructured off- road environments remains a longstanding challenge. In the absence of structured roads and in the presence of uneven terrain, vegetation, and soil slopes, vehicles must rely on LiDAR–Camera fusion to identify stable and traversable roads. However, existing terrain perception methods largely remain at the level of semantic segmentation and struggle to cap- ture physical attributes such as surface roughness and load- bearing capacity. Meanwhile, constructing datasets annotated with accurate physical properties is prohibitively costly and inherently limited in class diversity, making it difficult to cover unseen terrains. To address these limitations, we propose an online ground bumpiness cost learning framework for off- road vehicles, which enables continuous and direct learning of terrain-specific bumpiness costs during operation without the need for manual annotation. The framework consists of four key components: (i) ground bumpiness cost computation, (ii) a lightweight multimodal terrain segmentation model, (iii) an instance-level incremental update strategy, and (iv) a bumpiness cost mapping module. Extensive experiments on the EV-56 vibroseis truck demonstrate that the proposed framework can finely discriminate terrains with varying bumpiness costs and incrementally estimate costs for previously unseen terrains, thereby providing strong support for safe and reliable off-road autonomous driving.

Index terms

Field Robots Robotics in Hazardous Fields Incremental Learning

Related papers