Risk-aware Integrated Task and Motion Planning for Versatile Snake Robots under Localization Failures

TL;DR

This work addresses robust navigation for snake robots operating in challenging environments where exteroceptive localization can fail due to near-ground sensing. It introduces BLISS-TAMP, which casts integrated task-and-motion planning under uncertainty as a convex MILP by reformulating the challenging CC-HPOMDP problem into a tractable optimization, while treating long-range planning at a 2.5-D level and handling high-DoF execution with a short-range controller. The approach yields significant practical benefits, achieving over an order of magnitude faster planning than state-of-the-art POMDP solvers and more than 50% better navigation-time optimality than classical two-stage planners, demonstrated in simulations and hardware experiments on the EELS snake robot. The results indicate broad applicability of convex, risk-aware TAMP for proprioceptive mobility with intermittent sensing, offering a scalable path toward reliable autonomous operation in extreme terrains.

Abstract

Snake robots enable mobility through extreme terrains and confined environments in terrestrial and space applications. However, robust perception and localization for snake robots remain an open challenge due to the proximity of the sensor payload to the ground coupled with a limited field of view. To address this issue, we propose Blind-motion with Intermittently Scheduled Scans (BLISS) which combines proprioception-only mobility with intermittent scans to be resilient against both localization failures and collision risks. BLISS is formulated as an integrated Task and Motion Planning (TAMP) problem that leads to a Chance-Constrained Hybrid Partially Observable Markov Decision Process (CC-HPOMDP), known to be computationally intractable due to the curse of history. Our novelty lies in reformulating CC-HPOMDP as a tractable, convex Mixed Integer Linear Program. This allows us to solve BLISS-TAMP significantly faster and jointly derive optimal task-motion plans. Simulations and hardware experiments on the EELS snake robot show our method achieves over an order of magnitude computational improvement compared to state-of-the-art POMDP planners and $>$ 50\% better navigation time optimality versus classical two-stage planners.

Figures (23)

• Figure 1: Blind-motion with Intermittently Scheduled Scans (BLISS).
• Figure 2: EELS' layered autonomy architecture: from Long-range TAMP (this work), Short-range planner, to High-DoF controllers.
• Figure 3: BLISS-TAMP introduces hyperbeliefs. After initial scanning, the state's mean depends on the observation. This transforms belief into a distribution over distributions for post-scan state prediction.
• Figure 4: Standard map
• Figure 5: Entrapped map