REVIEW 23 cited by
Parkour in the Wild: Learning a General and Extensible Agile Locomotion Policy Using Multi-expert Distillation and RL Fine-tuning
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
Parkour in the Wild: Learning a General and Extensible Agile Locomotion Policy Using Multi-expert Distillation and RL Fine-tuning
read the original abstract
Legged robots are well-suited for navigating terrains inaccessible to wheeled robots, making them ideal for applications in search and rescue or space exploration. However, current control methods often struggle to generalize across diverse, unstructured environments. This paper introduces a novel framework for agile locomotion of legged robots by combining multi-expert distillation with reinforcement learning (RL) fine-tuning to achieve robust generalization. Initially, terrain-specific expert policies are trained to develop specialized locomotion skills. These policies are then distilled into a unified foundation policy via the DAgger algorithm. The distilled policy is subsequently fine-tuned using RL on a broader terrain set, including real-world 3D scans. The framework allows further adaptation to new terrains through repeated fine-tuning. The proposed policy leverages depth images as exteroceptive inputs, enabling robust navigation across diverse, unstructured terrains. Experimental results demonstrate significant performance improvements over existing methods in synthesizing multi-terrain skills into a single controller. Deployment on the ANYmal D robot validates the policy's ability to navigate complex environments with agility and robustness, setting a new benchmark for legged robot locomotion.
Forward citations
Cited by 23 Pith papers
-
Light-Loco-Parkour: Versatile Perceptive Whole-Body Locomotion via Multi-Skill Distillation
A single neural-network policy, trained in simulation, makes a humanoid climb, vault, and traverse uneven terrain from onboard depth and a velocity command, with no skill labels or runtime motion graphs.
-
HumanoidArena: Benchmarking Egocentric Hierarchical Whole-body Learning
HumanoidArena is a new benchmark of 7 leg-critical HOI/HSI tasks that evaluates egocentric hierarchical whole-body policies in humanoids and finds performance is strongly conditioned on the low-level GMT used.
-
roto 2.0: The Robot Tactile Olympiad
roto 2.0 provides a standardized benchmark for end-to-end blind tactile RL on 16-24 DOF robots, with open-sourced baselines achieving 13 Baoding ball rotations in 10 seconds.
-
Extreme-RGMT: Continual Learning of Highly Dynamic Skills for Robust Generalist Humanoid Control
A two-stage continual-learning framework lets a generalist humanoid tracking policy acquire highly dynamic acrobatic skills while preserving its general-purpose motion capabilities.
-
EgoHTR: Egocentric 4D Demonstrations of Human Terrain Traversal
EgoHTR is a 55-sequence, 150k-frame egocentric 4D human-terrain dataset with a reconstruction pipeline, MoCap-validated benchmark, and perceptive locomotion policies deployed on a Unitree G1.
-
Learning Locomotion on Discrete Terrain via Minimal Proximity Sensing
Foot-mounted proximity sensors provide pre-contact feedback that, when integrated into RL, improves quadruped traversal robustness on discrete terrain with reliable sim-to-real transfer.
-
StairMaster: Learning to Conquer Risky Hollow Stairs for Agile Quadrupedal Robots
StairMaster trains an RL policy that lets a Unitree Go2 quadruped climb hollow stairs up to 55 degrees via zero-shot sim-to-real transfer using cross-attention, SRU memory, and active-perception rewards.
-
StairMaster: Learning to Conquer Risky Hollow Stairs for Agile Quadrupedal Robots
A three-stage RL framework with cross-attention, spatial memory, and depth-noise augmentation lets a Unitree Go2 climb 55° hollow stairs zero-shot from simulation.
-
SWAP: Symmetric Equivariant World-Model for Agile Robot Parkour
SWAP embeds symmetry equivariance into world models and policies, enabling a quadruped to leap 2.13m gaps and climb 1.63m platforms with robust generalization to mirrored and outdoor terrains.
-
TAGA: Terrain-aware Active Gaze Learning for Generalizable Agile Humanoid Locomotion
TAGA learns terrain-aware active gaze behaviors for humanoid robots via RL alone, enabling generalizable locomotion with 1.2m real-world gap traversal.
-
Online World Modeling Enables Real-World Inverse Reinforcement Learning from Observation
MPAIL2 demonstrates real-world manipulation learning from observation alone, without rewards or action labels, plus positive online transfer.
-
Perceptive Humanoid Parkour: Chaining Dynamic Human Skills via Motion Matching
A modular system uses motion matching to compose long-horizon human skill chains, trains RL experts, and distills them into a depth-based policy that lets a Unitree G1 humanoid autonomously climb, vault, and roll over...
-
Isaac Lab: A GPU-Accelerated Simulation Framework for Multi-Modal Robot Learning
Isaac Lab is a unified GPU-native platform combining high-fidelity physics, photorealistic rendering, multi-frequency sensors, domain randomization, and learning pipelines for scalable multi-modal robot policy training.
-
Continual-RL for Generalization in Autonomous Racing on the RoboRacer Platform
SAC plus Continual Backpropagation, trained only on real multi-track data, fine-tunes in ~15 minutes on an unseen lower-friction RoboRacer track and outperforms MAP and MPC.
-
Learning Locomotion on Discrete Terrain via Minimal Proximity Sensing
Foot-mounted infrared proximity sensors supply pre-contact signals that, when incorporated into RL, improve quadrupedal traversal of discrete terrain such as gaps and stepping stones.
-
LadderMan: Learning Humanoid Perceptive Ladder Climbing
A hybrid motion-tracking and imitation-reinforcement pipeline produces a depth-based visuomotor policy that lets humanoids climb varied ladders zero-shot on hardware and perform teleoperated manipulation while climbing.
-
CoRe-MoE: Contrastive Reweighted Mixture of Experts for Multi-Terrain Humanoid Locomotion with Gait Adaptation
CoRe-MoE uses a two-stage RL framework with contrastive reweighting in a Mixture-of-Experts architecture to enable gait transitions and multi-terrain adaptation for humanoid locomotion.
-
SSR: Scaling Surefooted and Symmetric Humanoid Traversal to the Open World
SSR is an end-to-end vision-based framework for humanoid traversal that learns imagined foothold guidance, equivariant latent-space symmetry augmentation, and terrain-specific multi-discriminator motion priors to enab...
-
DPL: Depth-only Perceptive Humanoid Locomotion via Realistic Depth Synthesis and Cross-Attention Terrain Reconstruction
Combining a blind-backbone policy, cross-attention terrain reconstruction from depth plus proprioception, and realistic synthetic depth with noise enables depth-only full-sized humanoid locomotion over stairs, slopes,...
-
HERMES: Human-to-Robot Embodied Learning from Multi-Source Motion Data for Mobile Dexterous Manipulation
HERMES converts a single human motion demonstration into a deployable mobile bimanual dexterous manipulation policy, using RL, depth-image distillation, and closed-loop PnP pose refinement.
-
Efficient On-policy Visual-RL via Stochastic Decoupled Policy Gradient
SDPG is a new on-policy visual RL algorithm that estimates gradients via stochastic perturbations of rollouts, achieving faster training and lower memory use than baselines on visual MuJoCo tasks while adding new robo...
-
Evaluation of an Actuated Spine in Agile Quadruped Locomotion
Adding an actuated sagittal spine to a simulated quadruped increases agility and allows it to clear higher obstacles, steeper slopes, and tighter passages than the rigid-spine baseline.
-
Quadruped Parkour Learning: Sparsely Gated Mixture of Experts with Visual Input
Sparsely gated MoE policies double the success rate of a real Unitree Go2 quadruped on large-obstacle parkour versus matched-active-parameter MLP baselines while cutting inference time compared with a scaled-up MLP.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.