Escape Without a Recipe: Maneuver-Agnostic Mars Rover Recovery from Granular Entrapment

Meraj Hossain Promit; Chandak Chakma; Md Jubair Ahmed Sourov; Sejuti Rahman

cs.ROOct 17, 2026

Escape Without a Recipe: Maneuver-Agnostic Mars Rover Recovery from Granular Entrapment

Authors: Meraj Hossain Promit, Chandak Chakma, Md Jubair Ahmed Sourov, Sejuti Rahman

Organizations: Department of Robotics & Mechatronics Engineering, University of Dhaka, Bangladesh. · Department of AI & Robotics, New Uzbekistan University, Uzbekistan.

Abstract

Wheeled planetary rovers immobilised in loose regolith have no principled recovery controller: scripted maneuvers are terrain-specific, and naive driving deepens entrapment. We introduce a recurrent policy that learns to escape without a predefined maneuver. The framework rests on a verified trap: in a continuum material point method (MPM) simulator we identify a configuration at sinkage-to-radius ratio z/r = 2.0 where the strongest of ten scripted controllers succeeds in only 7/32 trials, yet the regolith retains traction and continues to shear, confirming granular entrapment rather than a numerical artifact. A frozen entrapment detector, trained on 81,771 windows spanning three materials and six sinkage levels, supplies a three-state belief to the policy observation. A separate simulator mobility latch gates training rewards. Across five training seeds, mean action evaluation achieves 62.5% escape (range 28.1-81.3%; 32 trials per seed) versus 7/32 for the strongest scripted controller and 0/32 for action noise. A numerical audit identifies a coupling-buffer fault that silently discarded 9.7% of contact nodes at 16 parallel environments. Results establish recovery in the tested simulation regime; physical transfer remains unverified.

Figures & tables

Fig. 2: Evaluated recovery loop. A frozen convolutional neural network (CNN)–gated recurrent unit (GRU) maps motion history to normal, sinking, and entrapped probabilities. These augment 29 simulation channels; the normalized 32-D actor input produces ten commands, filtered with $\alpha=0.4$ at 25 Hz. The training-only critic receives eight additional privileged features. Normalization and scaling are omitted; traces and probabilities are schematic.

Fig. 3: Training pipeline and proposed navigator interface. Top: a competence-gated burial curriculum supplies 16 parallel MPM environments; proximal policy optimization (PPO) trains the actor and critic with random network distillation (RND). Bottom: proposed navigation/recovery handoff, not evaluated. Experiments assess the recovery policy alone.

Input: frozen detector $D$ , actor $\pi_{\theta}$ , critic $V_{\phi}$ ;
16 environments, 100k control steps/environment.
1	Initialize observations, recurrent and filter states.
2	while the control-step budget remains do
3	Collect 256 steps/env: $b_{t}{=}D(\text{sensor history})$ ; normalize $o_{t}{=}[x_{t};b_{t}]\in\mathbb{R}^{32}$ .
4	Sample $a_{t}{\sim}\mathcal{N}(\mu_{t},\operatorname{diag}\sigma_{a}^{2})$ ; filter $u_{t}{=}0.4a_{t}{+}0.6u_{t-1}$ at 25 Hz.
5	Receive reward ( 2 ), next obs, done flags; env updates latch.

Table 3

Fig. 4: Depth screening. (a) Recoveries/16 for eight scripted controllers and action noise. (b) Slip-derived grip. At $z/r=2.0$ all arms except paddling score zero, yet grip stays nonzero ( $0.05$ – $0.22$ ): the regolith still carries traction, confirming granular rather than numerical entrapment.

Fig. 5: Training-seed and action-mode variability at

z/r=2.0

. Points show recoveries/32, with pointwise 95% Wilson intervals. Dashed lines mark the best scripted rate (

7/32

). Rows share training seeds, not paired trajectories.

	$z/r=1.8$		$z/r=2.0$
Training variant	Mean	Sampled	Mean	Sampled
Full policy	32/32	27/32	23/32	14/32
Without belief	31/32	29/32	19/32	11/32
Raw- $P_{\rm ent}$ gate	16/32	11/32	12/32	5/32

TABLE I: Detector-belief ablation. Recoveries/32 per cell in a separate single-training-seed comparison.

Robot Failure Recovery

Robot failure recovery studies how robots detect, diagnose, and respond to errors, damage, and unexpected physical interactions.

See all 72 papers

Towards Low-Gravity Planetary Exploration using Reinforcement Learning for Walking, Jumping, and In-flight Attitude Control
May 23, 2026Jørgen Anker Olsen, Kostas AlexisRobotic Control Space Robotics
The authors are with the Autonomous Robots Lab, NTNU, O.S. Bragstads Plass 2D, 7034, Trondheim, Norway.
Learning All-Terrain Locomotion for a Planetary Rover with Actively Articulated Suspension
Jun 5, 2026Arthur Bouton, Tristan D. Hasseler, Michael Paton +5Feedback Control Off-Road Robot Navigation
Jet Propulsion Laboratory, California Institute of Technology, Pasadena, CA 91109 USA · Center for Autonomy, University of Texas at Austin, Austin, TX 78712 USA · Space Systems Laboratory, University of Maryland, MD 20742 USA
Locomotion analysis of a quadruped interacting with the lunar granular surface
Jun 9, 2026Yash J VyasGranular Terrain Locomotion Robotics Simulation
Department of Industrial Engineering, University of Padua

May 23, 2026J/K move · Enter open · S save

Towards Low-Gravity Planetary Exploration using Reinforcement Learning for Walking, Jumping, and In-flight Attitude Control

Jørgen Anker Olsen, Kostas Alexis

The authors are with the Autonomous Robots Lab, NTNU, O.S. Bragstads Plass 2D, 7034, Trondheim, Norway.

Robotic Control Space Robotics Robot Locomotion RL for Legged Locomotion Quadruped Locomotion

Open paper AskPDF arXiv

This paper presents reinforcement learning (RL) policies for dynamic quadrupedal locomotion in planetary exploration scenarios. Building on a taskoptimized quadruped with a 5-bar leg design, we develop RL policies for walking, vertical jumping, forward jumping, and in-flight attitude control, explicitly tailored to the reduced gravity on Mars. These policies jointly enable such robots to overcome obstacles larger than themselves through coordinated jumping and precise in-flight reorientation for safe landings. We demonstrate Sim2Real transfer of the attitude control policy on the Olympus quadruped through single-axis reorientation tests, while all locomotion policies are validated in simulation. A complete Mars exploration mission scenario demonstrates coordinated policy deployment across challenging terrain. Experimental results show 90° attitude reorientation in 2.6 seconds, with simulations demonstrating 3.1 meter vertical jumps and 3.9 meter forward jumps under Martian gravity conditions. - Supplementary video: https://www.youtube.com/watch?v=qlSJ3P87A4A

Abstract

Robot Failure Recovery

Explore similar work

Towards Low-Gravity Planetary Exploration using Reinforcement Learning for Walking, Jumping, and In-flight Attitude Control

Learning All-Terrain Locomotion for a Planetary Rover with Actively Articulated Suspension

Locomotion analysis of a quadruped interacting with the lunar granular surface