Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

SlipNet: Enhancing Slip Cost Mapping for Autonomous Navigation on Heterogeneous and Deformable Terrains

Mubarak Yakubu, Yahya Zweiri, Ahmad Abubakar, Rana Azzam, Ruqayya Alhammadi, Lakmal Seneviratne

TL;DR

SlipNet addresses wheel-slip prediction on deformable, heterogeneous terrains for autonomous rovers without relying on prior terrain labels. It fuses a DeepLab v3+ terrain segmentation network with a Vision Transformer–based SlipNet that dynamically updates slip predictions per terrain segment, producing a Slip Cost Map for navigation. Trained on a large synthetic dataset generated with Vortex Studio, SlipNet outperforms TerrainNet in MAE across unseen terrains and path scenarios, demonstrating improved traction-aware autonomous navigation. The approach enables real-time, adaptive terrain analysis and slip prediction, enhancing robustness and safety for planetary exploration missions.

Abstract

Autonomous space rovers face significant challenges when navigating deformable and heterogeneous terrains due to variability in soil properties, which can lead to severe wheel slip, compromising navigation efficiency and increasing the risk of entrapment. To address this problem, we introduce SlipNet, a novel approach for predicting wheel slip in segmented regions of diverse terrain surfaces without relying on prior terrain classification. SlipNet employs dynamic terrain segmentation and slip assignment techniques on previously unseen data, enhancing rover navigation capabilities in uncertain environments. We developed a synthetic data generation framework using the high-fidelity Vortex Studio simulator to create realistic datasets that replicate a wide range of deformable terrain conditions for training and evaluation. Extensive simulation results demonstrate that our model, combining DeepLab v3+ with SlipNet, significantly outperforms the state-of-the-art TerrainNet method, achieving lower mean absolute error (MAE) across five distinct terrain samples. These findings highlight the effectiveness of SlipNet in improving rover navigation in challenging terrains.

SlipNet: Enhancing Slip Cost Mapping for Autonomous Navigation on Heterogeneous and Deformable Terrains

TL;DR

SlipNet addresses wheel-slip prediction on deformable, heterogeneous terrains for autonomous rovers without relying on prior terrain labels. It fuses a DeepLab v3+ terrain segmentation network with a Vision Transformer–based SlipNet that dynamically updates slip predictions per terrain segment, producing a Slip Cost Map for navigation. Trained on a large synthetic dataset generated with Vortex Studio, SlipNet outperforms TerrainNet in MAE across unseen terrains and path scenarios, demonstrating improved traction-aware autonomous navigation. The approach enables real-time, adaptive terrain analysis and slip prediction, enhancing robustness and safety for planetary exploration missions.

Abstract

Autonomous space rovers face significant challenges when navigating deformable and heterogeneous terrains due to variability in soil properties, which can lead to severe wheel slip, compromising navigation efficiency and increasing the risk of entrapment. To address this problem, we introduce SlipNet, a novel approach for predicting wheel slip in segmented regions of diverse terrain surfaces without relying on prior terrain classification. SlipNet employs dynamic terrain segmentation and slip assignment techniques on previously unseen data, enhancing rover navigation capabilities in uncertain environments. We developed a synthetic data generation framework using the high-fidelity Vortex Studio simulator to create realistic datasets that replicate a wide range of deformable terrain conditions for training and evaluation. Extensive simulation results demonstrate that our model, combining DeepLab v3+ with SlipNet, significantly outperforms the state-of-the-art TerrainNet method, achieving lower mean absolute error (MAE) across five distinct terrain samples. These findings highlight the effectiveness of SlipNet in improving rover navigation in challenging terrains.
Paper Structure (15 sections, 8 equations, 6 figures, 2 tables)

This paper contains 15 sections, 8 equations, 6 figures, 2 tables.

Table of Contents

  1. INTRODUCTION
  2. Related Work
  3. Current practice on Mars
  4. Traversability from Semantics
  5. Self-supervision
  6. Methodology
  7. Background on Slip
  8. Slip Risk Module
  9. Data Generation
  10. Model Architecture
  11. Terrain Semantic Segmentation
  12. SlipNet
  13. Results
  14. Conclusions
  15. Acknowledgements.

Figures (6)

  • Figure 1: Concept of the proposed SlipNet that takes as input multiple terrain semantic images and in-situ wheel slip and speed measurements. By utilizing past experiences on various terrain surfaces, the proposed method enhances prediction accuracy on potentially hazardous terrains in new environments, where training data is limited or not available.
  • Figure 2: Slip versus Speed dataset generated for eight different terrain types
  • Figure 3: Our proposed scheme is constructed as a dual-network architecture including a segmentation network and the SlipNet
  • Figure 4: Overview of the encoder-decoder architecture of the DeepLab v3+
  • Figure 5: Qualitative comparison between different semantic segmentation networks combination with SlipNet and the baseline, in five different terrain samples. DeepLab v3+ is the best candidate in Slip cost map due to weak supervision which is necessary for segmenting terrain texture not included in the annotation mask
  • ...and 1 more figures