Organizations: Department of Artificial Intelligence, Korea University, Seoul, Republic of Korea. · ROBOTIS · Department of Statistics, Korea University, Seoul, Republic of Korea.
Humanoid motion-tracking policies can reproduce diverse behaviors, yet connecting independently acquired motions into a continuous sequence remains challenging. Interpolation connects boundary poses without considering whether their relative position, heading, and timing provide suitable conditions for a transition. A generative motion prior can produce richer motion, but poorly arranged boundaries may cause it to generate excessive locomotion, turning, or another unintended action. We present Spatio-Temporal Alignment with a Motion Prior (STAMP), a framework for connecting independently acquired humanoid motions through a Robot-to-Human-to-Robot process. STAMP uses human motion space to draw on rich motion priors and reuse transition knowledge across robot embodiments. Its key component is a learned alignment model that determines the boundary conditions for motion inbetweening. Given the two motions retargeted to a human rig, the alignment model ranks candidate positions, headings, and transition durations for the following motion. A human motion prior generates the transition under the selected alignment. The completed sequence is retargeted to the robot and used as a reference to train a motion-tracking policy. We evaluate the tracking success of the generated transitions on UnitreeG1 andROBOTIS K1. Physical robot demonstrations further show that STAMP connects independently acquired motions into long-horizon sequences. Project page: https://childtoy.github.io/STAMP-project
Figures & tables
Fig. 1 : Overview of STAMP. Robot motions are transferred to a human rig. The alignment model predicts a transition cost for each candidate position, heading, and duration. The lowest-cost candidate is selected to condition motion inbetweening in human space. The completed sequence is retargeted to the robot and used as a motion-tracking reference.
Term
Definition
Weight
Eroot
Q(p¨r;3)+0.25Q(p...r;25)
1.0
Ebody
Q(p¨b;6)+0.25Q(p...b;50)
0.6
Efoot
Q(p¨f;5)+0.25Q(p...f;45)
0.8
Estance
Q(s˙;1)+0.25Q(s¨;6)
0.6
Eyaw
Q(ψ˙;π)+0.5Q(ψ¨;2π)
0.5
+0.1Q(ψ...;8π)
TABLE I: Kinematic supervision costs and combination weights. Normalization scales use the corresponding SI units.
Method
Success Rate (%) ↑
Evel (m/s) ↓
Eacc (rad/s 2 ) ↓
Root/heading interpolation
69.59
0.2581
13.57
MotionBricks
56.91
0.3433
17.67
Kimodo †
50.69
0.5293
18.98
STAMP
80.69
0.2289
12.43
TABLE II: Unitree G1 transition tracking over 18 motion pairs. Kimodo † denotes the Kimodo-based inbetweening model.
Method
Success Rate (%) ↑
Evel (m/s) ↓
Eacc (rad/s 2 ) ↓
Root/heading interpolation
84.70
0.1964
11.74
Kimodo †
64.55
0.5303
16.50
STAMP
87.52
0.1860
11.12
TABLE III: ROBOTIS K1 transition tracking over 18 motion pairs. Kimodo † denotes the Kimodo-based inbetweening model.
Robot
Candidate
Success Rate (%) ↑
Evel (m/s) ↓
Eacc (rad/s 2 ) ↓
Unitree G1
Rank 1
80.69
0.2289
12.43
Rank 100
71.33
0.3001
15.97
ROBOTIS K1
Rank 1
87.52
0.1860
11.12
Rank 100
71.90
0.2875
14.34
TABLE IV: Tracking performance at alignment ranks 1 and 100. Rank 1 has the lowest predicted transition cost.
Fig. 2 : Running-to-punching transitions on ROBOTIS K1 using interpolation (top) and STAMP (bottom), progressing from right to left. STAMP places the punching motion farther forward, providing space for a connecting step during the transition.
Fig. 3 : Motor-current response during a transition from looking around to making a stop gesture on ROBOTIS K1. Both methods use a 1.5 s transition. Lines and shading indicate the mean and ±1 standard deviation across five executions of the RMS signal over 23 motors. Time zero denotes transition onset. Values are in driver-reported units.
Fig. 4 : Long-horizon compositions of the same four traffic-control motions executed on the ROBOTIS K1. Each row shows a different composition. STAMP selects the spatial and temporal alignment and generates the transitions between consecutive motions.
Robot motion generation creates feasible, task-directed trajectories for robots to move, manipulate objects, and interact with people and environments.
Retargeting human motion to humanoid robots is critical for teleoperation, imitation learning and human-robot interaction. However, it remains challenging because of substantial morphological discrepancies between humans and robots, including differences in skeletal topology, limb proportions and degrees of freedom, as well as the scarcity of paired motion data. This paper presents Human2Humanoid, an unsupervised motion retargeting framework that transfers human motions to humanoid robot behaviors with high fidelity. To bridge the domain gap under unpaired data, we adopt a CycleGAN-based architecture equipped with a skeleton-aware graph convolutional network to capture topology-dependent motion features. To address cross-domain scale mismatches, we introduce a morphology-invariant end-effector consistency loss that aligns normalized end-effector trajectories to preserve motion semantics across embodiments. To improve physical plausibility and reduce contact artifacts, we impose explicit physics-aware feasibility constraints to encourage reproduction of the contact patterns in the source motion. Experimental results show that the proposed method successfully retargets human motion to the Unitree G1 humanoid robot without paired data, and outperforms existing methods in both downstream controllability and physical feasibility.
Tianchen Huang, Feiyang Yuan, Junchi Gu +5
Institute of Humanoid Robots, Department of Precision Machinery and Precision Instrumentation, University of Science and Technology of China, Hefei, Anhui 230026, China
General-purpose humanoid trackers can execute diverse references, but their zero-shot coverage depends on large embodied corpora that are costly to extend. Text-to-motion generators offer scalable supervision, yet models trained on human motion or retargeted data inherit a gap between kinematic plausibility and robot executability. Existing one-way pipelines fix either the generated corpus or the reward tracker. We introduce GenTrack, an online generator--tracker framework that alternates execution-grounded, group-relative generator alignment with tracker training on newly generated references; anchoring and rehearsal constrain drift. On Unitree G1, we evaluate GenTrack with ProtoMotions and SONIC backbones across three zero-shot tracking splits including public AMASS and LAFAN benchmarks, and a private out-of-distribution test set of 1,024 prompt-motion pairs in the wild. The online co-training strategy consistently produces generators that output more robot-executable motions with strong semantic alignment, and trackers with markedly broader zero-shot coverage and improved tracking accuracy, especially on out-of-distribution references. These results demonstrate that joint online post-training effectively narrows the executability gap between retargeted references and robot-native motion, advancing zero-shot humanoid control without additional data collection and beyond the limitations of a static reference pool.
Zeyu Ling, Xinyao Yu, Renye Yan +4
Zhejiang University · Peking University · Tencent +1
While velocity-commanded RL policies are effective for controlling quadrupeds, they are a limited interface for humanoids that need to perform complex whole-body motions. Furthermore, RL tracking policies for whole-body motions typically have limited information about the desired future motion, preventing them from performing controlled deviations to achieve long-term task success; such information is particularly relevant when a controller must prepare for transitions between behaviors. We address these challenges using concise motion representations that can be generated by a high-level controller, given a behavior label or text command. These representations can be fed directly into a tracking policy to provide a detailed motion reference. Our method exhibits a high degree of flexibility in performing real-world robot motions such as dancing, fighting, or walking based on conditional information. At matched capacity, our adopted stochastic motion tokenizer reconstructs motions from the Unitree G1 at a level that exceeds state-of-the-art approaches such as finite scalar quantization (FSQ) and the vector-quantized autoencoder (VQ-VAE). The resulting discrete interface conditions whole-body tracking policies that execute generated dance, fight, and walk motions lasting 16.8 s each and a 35 s fight-to-dance sequence on a physical Unitree G1.
Michael Drolet, Oleg Arenz, Jan Peters
Technical University of Darmstadt · German Research Center for AI (DFKI) · Robotics Institute Germany (RIG)