PRIOR: Perceptive Learning for Humanoid Locomotion with Reference Gait Priors

Chenxi Han; Shilu He; Yi Cheng; Linqi Ye; Houde Liu

cs.ROMar 19, 2026

PRIOR: Perceptive Learning for Humanoid Locomotion with Reference Gait Priors

Authors: Chenxi Han, Shilu He, Yi Cheng, Linqi Ye, Houde Liu

Organizations: Tsinghua University · ZERITH Robotics · Shanghai University

Abstract

Training perceptive humanoid locomotion policies that traverse complex terrains with natural gaits remains an open challenge, typically demanding multi-stage training pipelines, adversarial objectives, or extensive real-world calibration. We present PRIOR, an efficient and reproducible framework built on Isaac Lab that achieves robust terrain traversal with human-like gaits through a simple yet effective design: (i) a parametric gait generator that supplies stable reference trajectories derived from motion capture without adversarial training, (ii) a GRU-based state estimator that infers terrain geometry directly from egocentric depth images via self-supervised heightmap reconstruction, and (iii) terrain-adaptive footstep rewards that guide foot placement toward traversable regions. Through systematic analysis of depth image resolution trade-offs, we identify configurations that maximize terrain fidelity under real-time constraints, substantially reducing perceptual overhead without degrading traversal performance. Comprehensive experiments across terrains of varying difficulty-including stairs, boxes, and gaps-demonstrate that each component yields complementary and essential performance gains, with the full framework achieving a 100% traversal success rate. We will open-source the complete PRIOR framework, including the training pipeline, parametric gait generator, and evaluation benchmarks, to serve as a reproducible foundation for humanoid locomotion research on Isaac Lab.

Figures & tables

Fig. 1 : The proposed PRIOR framework was simulated and demonstrated on the ZERITH Z1 model. (A)–(D) illustrate the robot traversing four representative terrain types.

Fig. 2 : Overview of the proposed PRIOR framework. The framework comprises three components: (a) Asymmetric actor-critic architecture for reinforcement learning. (b) State Estimator (yellow): Fuses multimodal latent features for policy driving and performs self-supervised regression for velocity estimation, terrain reconstruction, and state prediction. (c) Reference Gait Generator (blue): Synthesizes physics-consistent reference trajectories via phase normalization and velocity-driven weighted interpolation, while constraining locomotion through gait-aware rewards.

Reward	Description	Weight ( $w_{i}$ )
$r_{\text{air}}$	Promotes gait rhythm.	1.25
$r_{\text{slide}}$	Minimizes ground slipping.	-0.10
$r_{\text{dbl-air}}$	Penalizes walking on one leg.	-1.00
$r_{\text{swing}}$	Ensures leg lift height.	-20.0
$r_{\text{stumble}}$	Prevents foot-obstacle tripping.	-30.0
$r_{\text{edge}}^{L/R}$	Encourages safe foot placement.	-2.00

TABLE I : LANDING STATE REWARD COMPONENTS

Reward	Equation ( $e_{i}$ )	Weight ( $w_{i}$ )
$r_{\text{pos}}$	$\\|\boldsymbol{\theta}-\boldsymbol{\theta}_{d}(\phi)\\|^{2}$	0.10
$r_{\text{vel}}$	$\\|\mathbf{v}_{b}-\mathbf{v}\\|^{2}$	0.05
$r_{\Delta}$	$\\|\Delta\boldsymbol{\theta}-\Delta\boldsymbol{\theta}_{d}(\phi)\\|_{1}$	0.05
$r_{\text{ankle}}$	$\sum_{j\in\{L,R\}}\left(\theta_{j}-\theta_{d,j}(\phi)\right)^{2}$	0.05

TABLE II : GAIT-AWARE REWARD COMPONENTS

GPU	Max $N_{\mathrm{env}}$	Storage	VRAM
4090 (24G)	512	GPU	$\sim$ 24G
4090 (24G)	1024	CPU	$\sim$ 24G
4090 (48G)	1536	CPU	$\sim$ 48G

TABLE III : PARALLEL SCALE UNDER DIFFERENT MEMORY STRATEGIES

Fig. 3 : Depth data flow: Images in the top and bottom rows represent samples from different parallel environments.

Panel of Figure 7 — Fig. 4 : Overview of the terrain curriculum for training. (A)–(D) represent distinct terrain types posing various physical challenges. Table (E) details the parameter ranges and the distribution (weight) for each terrain type.

Parameter	Randomization range	Unit
Base payload	$[-5.0,5.0]$	kg
Link mass factor	$[0.8,1.2]$	–
Center of mass shift	$[-0.15,0.15]$	m
Friction coefficient	$[0.2,1.5]$	–
$K_{p}$ factor	$[0.9,1.1]$	–
$K_{d}$ factor	$[0.9,1.1]$	–

TABLE IV : DOMAIN RANDOMIZATION RANGES

Method	Mean Level	Pyramid Stairs	Inverted Stairs	Boxes	Plane	Mean Reward
PRIOR (ours)	5.7533	1.0	1.0000	1.0000	1.0	26.3462
PRIOR w/o reference gait	5.7735	1.0	1.0000	1.0000	1.0	23.7233
PRIOR w/o $\hat{\mathbf{m}}_{t}$	5.7672	1.0	0.7734	1.0000	1.0	13.1775
PRIOR w/o $\mathbf{d}_{t}^{\mathrm{H2}}$	5.4627	1.0	0.3750	0.9687	1.0	10.1463
PRIOR with H1 = 6	5.7417	1.0	1.0000	1.0000	1.0	19.3234
PRIOR w/o landing state reward	5.7403	1.0	1.0000	1.0000	1.0	22.6262

TABLE V : ABLATION STUDY RESULTS OF THE PRIOR FRAMEWORK ON TERRAIN ADAPTABILITY

Fig. 5 : The upper panel illustrates the training curves for mean reward, while the lower panel displays the mean terrain level achieved over training iterations.

Fig. 6 : Comparison of foot-placement behavior with and without the landing state reward.

Explore similar work

Generate, Track, Improve: Perceptive Multi-Skill Humanoid Locomotion with RL-Fine-Tuned Motion Generators
Sep 25, 2026Zachary Olkin, William D. Compton, Aaron D. AmesHumanoid Robot Locomotion Robot Motion Generation
Department of Control and Dynamical Systems at the California Institute of Technology
Echo in the Steps: Learning Perceptive Humanoid Parkour with Gated Memory
Sep 24, 2026Ming-Ju Lee, Zizhuo Wang, Shaoting Zhu +3Legged Robotics Humanoid Robot Locomotion
Tsinghua University
TAGA: Terrain-aware Active Gaze Learning for Generalizable Agile Humanoid Locomotion
Jun 4, 2026Peizhuo Li, Hongyi Li, Mingfeng Fan +9Active Perception Humanoid Robot Locomotion
MarmotLab, National University of Singapore · Center of X-Mechanics, Zhejiang University · South China University of Technology

Sep 25, 2026J/K move · Enter open · S save

Generate, Track, Improve: Perceptive Multi-Skill Humanoid Locomotion with RL-Fine-Tuned Motion Generators

Zachary Olkin, William D. Compton, Aaron D. Ames

Department of Control and Dynamical Systems at the California Institute of Technology

Humanoid Robot Locomotion Robot Motion Generation Terrain-Aware Robot Locomotion RL for Legged Locomotion Humanoid Motion Tracking

Open paper AskPDF arXiv

General purpose humanoids require locomotion controllers that are multi-skill, perceptive, dynamic, and robust enough to go anywhere humans can. In this work, we present a two layer locomotion architecture: (1) a perceptive flow matching motion generator plans whole body trajectories from raw depth images while a (2) perceptive tracking policy trained with control-guided RL follows these motions. Both policies are trained on a library of terrain consistent motion clips created with dynamically optimized human data which yields both accurate velocity tracking and terrain consistent references. Our central contribution is a simple yet effective off-policy RL fine tuning loop that improves the motion generator. A structured search method is used with the generator to gather data for advantage weighted regression. This off-policy loop is much more sample efficient than on-policy residual fine tuning and improves terrain consistency on unseen geometries and skill compositions. We find that successful terrain traversals increased by up to 25 percentage points and skill selection improved by up to 80 percentage points. By using raw depth images to perceive the environment no odometry or height maps are needed, and outdoor deployment is easy. With two cameras, the policy can see terrain coming from further away and adjust its velocity regardless of the commanded speed so it can traverse the terrain. A single policy pair enables a Unitree G1 humanoid to walk, run, stand, jump on and off of boxes, and traverse stairs in outdoor environments. Project page: https://zolkin1.github.io/generate-track-improve/

Abstract

Explore similar work

Generate, Track, Improve: Perceptive Multi-Skill Humanoid Locomotion with RL-Fine-Tuned Motion Generators

Echo in the Steps: Learning Perceptive Humanoid Parkour with Gated Memory

TAGA: Terrain-aware Active Gaze Learning for Generalizable Agile Humanoid Locomotion