Whole-Body Motion Guided Reference Generation for Uncertainty-Aware Compliant Assistance

Muyuan Ma; Long Cheng; Lijun Han; Haotian Zhai; Xinpan Meng; Xiuze Xia; Houcheng Li; Zeyu Liu

cs.ROOct 22, 2026

Whole-Body Motion Guided Reference Generation for Uncertainty-Aware Compliant Assistance

Authors: Muyuan Ma, Long Cheng, Lijun Han, Haotian Zhai, Xinpan Meng, Xiuze Xia, Houcheng Li, Zeyu Liu

Abstract

Exoskeletons have shown great potential for augmenting human physical capabilities, yet generating reliable assistance remains challenging because human motion patterns and interaction requirements vary substantially in practice. To address this problem, a dual-loop assistance strategy is proposed to couple multimodal exoskeleton trajectory generation with variable admittance control, where trajectory variance links the outer-loop generator and the inner-loop controller. The generator synthesizes future exoskeleton trajectories through denoising, primarily using whole-body human motion and exoskeleton proprioception, with task descriptions as auxiliary semantic cues. To reduce cross-modal representation mismatch, a three-stage training pipeline is introduced to progressively optimize the human motion encoder, the exoskeleton proprioception encoder, and the trajectory generator. In the inner loop, multiple future trajectories are sampled from the generator to estimate the trajectory mean and variance. The mean is used as the control reference, while the variance provides an uncertainty measure for prediction reliability. This measure is further incorporated into a stiffness regulation law, enabling compliant assistance with input-to-state stability under bounded lumped interaction input. Experimental results demonstrate that the proposed method significantly reduces muscle activation levels compared with standard baselines, while also decreasing users' metabolic energy expenditure relative to the unassisted condition.

Figures & tables

Fig. 1: Overview of the proposed dual-loop framework with uncertainty coupling for multimodal exoskeleton assistance.

Fig. 2: Architecture of the proposed multimodal exoskeleton trajectory generator.

Fig. 3: Three-stage training pipeline for multimodal exoskeleton trajectory generation.

Fig. 4: Simulation framework for human–exoskeleton interaction data generation.

Fig. 5: Motion tasks used for simulated human–exoskeleton interaction data generation.

Panel of Figure 6 — Fig. 6: Elbow exoskeleton platform and geometric illustration for gravity compensation. (a) Elbow exoskeleton with the interaction force interface and AK60-6 actuator. (b) Geometric illustration for the gravity compensation algorithm.

ID	Task	ID	Task
Task 1	Lift and carry a box from the right-side floor to the table in front	Task 8	OK gesture
Task 2	Move a box horizontally between floor positions	Task 9	Stop gesture
Task 3	Carry a box while walking forward	Task 10	Quiet gesture
Task 4	Lift a box from the table to the floor	Task 11	Wave gesture
Task 5	Lift a box from the floor to the table	Task 12	Salute gesture
Task 6	Repetitive box lifting and placing	Task 13	Six-box shelf stacking

TABLE I: Task list of the real exoskeleton dataset.

Training Scheme	Exoskeleton MAE (rad) $\downarrow$	Exoskeleton MSE (rad 2 ) $\downarrow$	Exoskeleton RMSE (rad) $\downarrow$
Stage 3 only	$0.2259\pm 0.1259$	$0.1482\pm 0.1595$	$0.3441\pm 0.1727$
Stage 2 + Stage 3	$0.1835\pm 0.1137$	$0.1148\pm 0.1373$	$0.2914\pm 0.1729$
Stage 1 + Stage 2 + Stage 3	$\mathbf{0.1666\pm 0.1039}$	$\mathbf{0.1002\pm 0.1296}$	$\mathbf{0.2741\pm 0.1582}$

TABLE II: Ablation results of the three-stage training pipeline. The results are reported as mean

\pm

standard deviation.

Conditional Inputs	Exoskeleton MAE (rad) $\downarrow$	Exoskeleton MSE (rad 2 ) $\downarrow$	Exoskeleton RMSE (rad) $\downarrow$
Full multimodal inputs	$\mathbf{0.1666\pm 0.1039}$	$\mathbf{0.1002\pm 0.1296}$	$\mathbf{0.2741\pm 0.1582}$
Without human motion	$0.1940\pm 0.1319$	$0.1300\pm 0.1715$	$0.3077\pm 0.1880$
Without task description	$0.1691\pm 0.1059$	$0.1009\pm 0.1265$	$0.2745\pm 0.1599$
Without human motion and task description	$0.1799\pm 0.1099$	$0.1122\pm 0.1450$	$0.2921\pm 0.1638$

TABLE III: Ablation results of multimodal conditional inputs. The results are reported as mean

\pm

standard deviation.

Fig. 7: Attention heatmap of task descriptions, human motion, and exoskeleton proprioception across 13 real exoskeleton tasks.

Fig. 8: Representative visualization of the role of human motion in future exoskeleton trajectory generation in Task 5. The upper plots show the observed exoskeleton joint positions and the future ground-truth and predicted trajectories, while the lower panels show the corresponding RGB frames and extracted human keypoints. The light-blue and light-red regions denote the observed input window and the future prediction window, respectively. Two phases are shown: (a) movement onset phase and (b) object approach phase.

Fig. 9: Experimental setup for comparative experiments. (a) Host computer. (b) USB-CAN adapter. (c) USB-ADC module. (d) Elbow exoskeleton [ 44 ] . (e) Power supply. (f) RGB-D camera.

Fig. 10: Representative illustrations and joint trajectory distributions of the 12 short-horizon tasks. The curves show the mean trajectories across seven subjects, and the shaded regions show the corresponding trajectory variations.

Task	Right BRR		Right BB		Right TB
Task	Proposed	Method in [ 18 ]	Proposed	Method in [ 18 ]	Proposed	Method in [ 18 ]
1	$\mathbf{15.95\pm 8.10}$	$20.51\pm 8.52$	$\mathbf{7.60\pm 3.73}^{*}$	$12.73\pm 6.28$	$\mathbf{11.36\pm 11.59}$	$11.95\pm 9.14$
2	$\mathbf{13.52\pm 5.95}$	$19.22\pm 4.31$	$\mathbf{6.70\pm 4.21}^{*}$	$13.29\pm 9.83$	$\mathbf{12.36\pm 8.97}$	$13.14\pm 9.22$
3	$\mathbf{16.00\pm 6.44}^{*}$	$22.83\pm 5.49$	$\mathbf{8.58\pm 4.26}^{*}$	$13.29\pm 9.73$	$\mathbf{11.80\pm 10.42}$	$13.42\pm 10.18$
4	$\mathbf{14.74\pm 7.04}^{*}$	$23.98\pm 10.26$	$\mathbf{8.24\pm 3.66}^{*}$	$15.60\pm 8.19$	$\mathbf{11.36\pm 10.43}$	$15.45\pm 10.97$
5	$\mathbf{14.11\pm 8.44}$	$18.99\pm 5.30$	$\mathbf{6.70\pm 3.89}^{*}$	$12.55\pm 10.51$	$\mathbf{9.89\pm 8.46}$	$13.42\pm 8.20$

TABLE IV: Muscle activation levels (%MVC) across subjects for 13 tasks under different assistance methods. The results are reported as mean

\pm

standard deviation.

Task level			Subject level
Task	Pearson $r$	Spearman $\rho$	Subject	Pearson $r$	Spearman $\rho$
1	0.6379	0.7346	1	0.6128	0.6801
2	0.6892	0.7737	2	0.5685	0.6427
3	0.6811	0.7297	3	0.5690	0.6813
4	0.7236	0.7400	4	0.5597	0.5498
5	0.7288	0.7737	5	0.6125	0.6636

TABLE V: Task and subject level correlations between trajectory variance and tracking error.

Fig. 11: Representative illustration and joint trajectory distribution of the long-horizon six-box carrying task. The task consists of six sequential box-carrying phases, and the lower panel shows the normalized joint trajectory distribution across seven subjects.

Fig. 12: Variable admittance control results for the dumbbell weightlifting task under 0 kg, 2 kg, and 5 kg loads. Curves and shaded regions denote the mean and standard deviation across seven subjects, respectively.

Load	Left BRR		Left BB		Left TB
Load	Method in [ 46 ]	Proposed	Method in [ 46 ]	Proposed	Method in [ 46 ]	Proposed
0 kg	$14.81\pm 6.18$	$\mathbf{10.30\pm 5.74}^{*}$	$14.93\pm 7.78$	$\mathbf{10.96\pm 3.93}^{*}$	$11.76\pm 10.17$	$\mathbf{9.00\pm 8.42}^{*}$
2 kg	$30.49\pm 15.18$	$\mathbf{21.29\pm 9.31}^{*}$	$24.89\pm 15.92$	$\mathbf{20.23\pm 10.74}^{+}$	$14.98\pm 16.17$	$\mathbf{14.31\pm 16.36}$
5 kg	$37.94\pm 15.41$	$\mathbf{32.87\pm 14.46}$	$37.42\pm 15.19$	$\mathbf{32.58\pm 13.45}$	$19.55\pm 13.17$	$\mathbf{16.92\pm 13.23}$

TABLE VI: Muscle activation levels (%MVC) across users under [ 46 ] and the proposed controller. The results are reported as mean

\pm

standard deviation.

Fig. 13: Energy expenditure comparison during repetitive 5 kg box lifting and placing. Curves and shaded regions denote the mean and standard deviation across seven subjects, respectively.

Fig. 14: Energy expenditure comparison in the last three sessions. Bars and error bars denote the mean and standard deviation across seven subjects, respectively. Statistical significance is marked as

*

for

p<0.05

and

**

for

p<0.01

Appendix figures & tables3 assets

Supplementary material from the paper’s appendix.

Appendix

Category	Parameter	Value
Data split	Training/test subjects	3/7, disjoint
RGB-D sensing	RGB-D camera	ZED 2i
Multimodal input	Input frequency	20 Hz
Input window	Human motion history $L_{h}$	20 frames
Input window	Proprioception history $L_{e}$	20 frames
Prediction	Future horizon $L_{p}$	40 frames

Appendix

TABLE VII: Implementation Parameters of the Proposed Framework

Task	Muscle	Reduction (%)	95% CI (%)	$d_{z}$	$p_{\mathrm{raw}}/p_{\mathrm{FDR}}$
1	Right BRR	$20\pm 39$	$[-16,\ 56]$	$0.509$	$0.2268/0.2474$
1	Right BB	$35\pm 24$	$[12,\ 57]$	$1.421$	$0.0094/0.0447^{*}$
1	Right TB	$11\pm 22$	$[-9,\ 32]$	$0.514$	$0.2229/0.2474$
1	Left BRR	$26\pm 31$	$[-3,\ 55]$	$0.839$	$0.0681/0.0908$
1	Left BB	$32\pm 19$	$[15,\ 50]$	$1.710$	$0.0040/0.0413^{*}$
1	Left TB	$24\pm 19$	$[6,\ 42]$	$1.235$	$0.0171/0.0447^{*}$

Appendix

TABLE VIII: Detailed statistical results for muscle activation comparison between the proposed method and [ 18 ] .

Load	Muscle	Reduction (%)	95% CI (%)	$d_{z}$	$p_{\mathrm{raw}}/p_{\mathrm{FDR}}$
0 kg	Left BRR	$32\pm 16$	$[17,\ 47]$	$1.942$	$0.0021/0.0110^{*}$
0 kg	Left BB	$20\pm 17$	$[5,\ 36]$	$1.193$	$0.0197/0.0443^{*}$
0 kg	Left TB	$24\pm 17$	$[9,\ 40]$	$1.438$	$0.0089/0.0267^{*}$
2 kg	Left BRR	$26\pm 14$	$[14,\ 39]$	$1.892$	$0.0024/0.0110^{*}$
2 kg	Left BB	$15\pm 14$	$[2,\ 28]$	$1.063$	$0.0307/0.0553$
2 kg	Left TB	$8\pm 15$	$[-5,\ 22]$	$0.554$	$0.1929/0.1929$

Appendix

TABLE IX: Detailed statistical results for the controller comparison under different external loads.

Assistive Robotics

Assistive robotics develops robots and wearable devices that support mobility, daily activities, rehabilitation, and care.

See all 207 papers

Ensuring Interaction Safety in Multitask Exoskeleton Control: A Simulation-Trained Variable Impedance Framework
Jun 4, 2026Muyuan Ma, Houcheng Li, Haotian Zhai +4Robotic Control Impedance Control
ExoTraj: A General Lower-limb Exoskeleton Assistance Policy for Complex Environments
Jun 15, 2026Xiao-Yin Liu, Guotao Li, Long Sun +2Upper-Limb Exoskeletons Assistive Robotics
The State Key Laboratory of Multimodal Artificial Intelligence Systems, Institute of Automation, Chinese Academy of Sciences, China · The School of Artificial Intelligence, University of Chinese Academy of Sciences, China · The School of Automation and Intelligence, Beijing Jiaotong University, China +1
Uncertainty-Aware Ankle Exoskeleton Control
Aug 28, 2025Fatima Mumtaza Tourk, Bishoy Galoaa, Sanat Shajan +3Uncertainty Quantification Ankle Exoskeleton Control
Department of Mechanical Engineering, Northeastern University, Boston, MA 02115 USA

Jun 4, 2026J/K move · Enter open · S save

Ensuring Interaction Safety in Multitask Exoskeleton Control: A Simulation-Trained Variable Impedance Framework

Muyuan Ma, Houcheng Li, Haotian Zhai, Lijun Han+3

Robotic Control Impedance Control Ankle Exoskeleton Control Language-Conditioned Robot Control

Open paper AskPDF arXiv

Wearable exoskeletons can augment human phys ical capabilities during complex activities. However, ensuring adaptation across diverse tasks while guaranteeing interaction safety remains a critical challenge. To address this, a simulation trained variable impedance control approach with stability guarantees is proposed. First, a simulation-based human exoskeleton motion data generation pipeline is established, utilizing Proximal Policy Optimization (PPO) to synthesize human muscle activations while the exoskeleton provides direct compensation for human biological joint torques. Subsequently, the generated dataset is used to train a dual modality policy that fuses semantic instructions with proprioceptive history, enabling the prediction of reference trajectories and variable impedance gains for nine different motion tasks. To guarantee safety, the network outputs are constrained by a stability criterion derived from Lyapunov stability theory, which bounds stiffness variations to ensure the asymptotic stability of the coupled system. Experimental results indicate that the proposed framework reduces metabolic cost in real-world scenarios com pared with standard baseline methods. These findings suggest the feasibility of the proposed framework for safe, multitask exoskeleton control.

Abstract

Assistive Robotics

Explore similar work

Ensuring Interaction Safety in Multitask Exoskeleton Control: A Simulation-Trained Variable Impedance Framework

ExoTraj: A General Lower-limb Exoskeleton Assistance Policy for Complex Environments

Uncertainty-Aware Ankle Exoskeleton Control