Attribution-Gated Joint Estimation of Intention and Capability for Assist-as-Needed Upper-Limb Robots

Pengcheng Li; Shuxiang Guo; Hanze Wang; Jun Leng; Bin Wang; Mingchao Ding

cs.ROOct 15, 2026

Attribution-Gated Joint Estimation of Intention and Capability for Assist-as-Needed Upper-Limb Robots

Authors: Pengcheng Li, Shuxiang Guo, Hanze Wang, Jun Leng, Bin Wang, Mingchao Ding

Organizations: Department of Electronic and Electrical Engineering, Southern University of Science and Technology, Shenzhen, China · Aerospace Center Hospital, School of Life Science, and the Key Laboratory of Convergence Medical Engineering System and Healthcare Technology, Ministry of Industry and Information Technology, Beijing Institute of Technology, Beijing, China · Department of Peripheral Vascular Intervention, Aerospace Center Hospital, School of Life Science, Beijing Institute of Technology, Beijing, 100081, China

Abstract

Assist-as-needed (AAN) control must distinguish between changes in intended motion and changes in effective voluntary torque capability, even though both can produce similar kinematic and interaction responses. A dual-measurand multimodal MHE-MPC architecture is developed as the base framework, in which the intended trajectory and the task-dependent effective voluntary torque capability are estimated online as distinct quantities, and both estimates are used to adapt robotic assistance. Building on this architecture, the ambiguity between the two update channels is formulated as a continuous credit-assignment problem and resolved by an attribution gate. A trajectory-change indicator is obtained from persistent corrective interaction, whereas a capability-change indicator is derived from the residual between normalized muscle activation and dynamics-reconstructed voluntary torque. Nine healthy participants completed factorial target-height-load trials and a pre-post fatigue cup-transport task. Without target-height input, the gated method generated correctly ordered Low-Middle-High references under every load: target changes were preferentially assigned to trajectory adaptation, whereas added load shifted adaptation toward the capability channel. It achieved the lowest RMSE in all nine conditions (0.017 rad versus 0.041 rad for ungated dual estimation and 0.127 rad for fixed-reference error-based AAN). Post-fatigue capability estimates decreased and assistance increased at all heights, supporting functional feasibility.

Figures & tables

Figure 1: Wearable upper-limb robot and control hardware. Cable-driven modules assist elbow and wrist motion, IMUs measure joint angles, and ROS 2 communicates with the embedded motor drivers.

Figure 2: Two-layer joint estimation and assistive-control method. Base MHE and MPC estimate capability and intended trajectory, respectively; both estimates determine assistance. The added gate is updated first:

1-\alpha

weights capability data and

\alpha

weights trajectory correction.

Figure 3: Experimental setup. (a) Participant wearing the upper-limb robot. (b) Three target heights used to vary movement intention. (c) Representative elbow trajectories. (d) Wrist-mounted loads used to vary task demand.

Figure 4: Experiment 2 functional cup-transport task. (a) A seated participant wearing the upper-limb robot reaches from the initial posture toward the cup. (b) Cup placement at the Low, Middle, and High target heights. (c) Representative joint-angle profiles for elbow flexion/extension, wrist flexion/extension, and forearm pronation/supination during one transport trial.

Figure 5: Continuous Experiment 1 responses from a representative recording, used to illustrate the controller behavior underlying the aggregate comparison. Signals are arranged by controller (columns) and signal group (rows). For each controller, consecutive raw trials in the Heavy-load block covering target heights Middle, Low, and High once each, retain that acquisition order, and were not selected based on response appearance. The top row overlays the measured elbow trajectory and the reference used by the controller; the repeated dotted trace is the same original, unupdated Middle-target nominal trajectory for all three target trials, whereas the dashed trace shows the resulting online-adapted reference. The next two rows show, on separate scales, logged attribution

\alpha

and the capability estimate

\hat{\tau}_{max}

in N

\cdot

m. The bottom row overlays normalized EMG (left axis) and commanded motor assistive torque (right axis, shown only on the rightmost panel).

Load	Height	Controller	Actual endpoint (rad)	Reference endpoint (rad)	Tracking RMSE (rad)	Mean attribution $\alpha$	Normalized EMG	Motor assistive torque (N m)
Free (0 kg)	Low	Error-based AAN	0.783 $\pm$ 0.037	0.980 $\pm$ 0.030	0.170 $\pm$ 0.028	0.500 $\pm$ 0.000	0.009 $\pm$ 0.004	0.151 $\pm$ 0.026
		Independent MPC	0.730 $\pm$ 0.028	0.690 $\pm$ 0.035	0.037 $\pm$ 0.013	0.500 $\pm$ 0.000	0.015 $\pm$ 0.011	0.007 $\pm$ 0.022
		Attribution-gated MPC	0.702 $\pm$ 0.034	0.686 $\pm$ 0.033	0.017 $\pm$ 0.014	0.573 $\pm$ 0.029	0.026 $\pm$ 0.028	0.004 $\pm$ 0.011
	Middle	Error-based AAN	1.018 $\pm$ 0.058	0.980 $\pm$ 0.030	0.067 $\pm$ 0.037	0.500 $\pm$ 0.000	0.015 $\pm$ 0.007	0.038 $\pm$ 0.047
		Independent MPC	0.986 $\pm$ 0.025	0.972 $\pm$ 0.026	0.034 $\pm$ 0.018	0.500 $\pm$ 0.000	0.018 $\pm$ 0.012	0.017 $\pm$ 0.023
		Attribution-gated MPC	1.002 $\pm$ 0.043	1.000 $\pm$ 0.048	0.013 $\pm$ 0.016	0.489 $\pm$ 0.034	0.026 $\pm$ 0.023	0.005 $\pm$ 0.010

Table 1: Experiment 1 outcomes.

Figure 6: Experiment 2 EMG (a) mean and (b) median frequency before and after fatigue, averaged across targets and repetitions within each participant.

Target	State	$\hat{\tau}_{max}$ (N m)	Mean assist (N m)	RMSE (rad)
Low	Fresh	47.240 $\pm$ 11.950	-0.005 $\pm$ 0.004	0.041 $\pm$ 0.022
Low	Post-fatigue	36.457 $\pm$ 20.231	0.244 $\pm$ 0.380	0.062 $\pm$ 0.012
Middle	Fresh	46.734 $\pm$ 12.298	0.009 $\pm$ 0.033	0.053 $\pm$ 0.023
Middle	Post-fatigue	36.260 $\pm$ 16.491	0.282 $\pm$ 0.325	0.060 $\pm$ 0.011
High	Fresh	45.268 $\pm$ 9.308	0.069 $\pm$ 0.077	0.056 $\pm$ 0.016
High	Post-fatigue	37.263 $\pm$ 20.503	0.369 $\pm$ 0.530	0.081 $\pm$ 0.015

Table 2: Experiment 2 outcomes by target height before and after fatigue (mean

\pm

sample standard deviation;

n=9

Figure 7: Representative Experiment 2 responses for one participant before and after fatigue. Each column contains three consecutive target-height trials in acquisition order; rows show elbow tracking,

\alpha

\hat{\tau}_{max}

, and normalized EMG/assistance.

Assistive Robotics

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

See all 208 papers

A Flow-Theoretic Multimodal Control Framework Fusing Mental and Physical Engagement for Upper-Limb Rehabilitation
Oct 15, 2026Jun Leng, Pengcheng Li, Hanze Wang +2Assistive Robotics Continuous Control
Department of Electronic and Electrical Engineering, Southern University of Science and Technology, Shenzhen 518055, China · Aerospace Center Hospital, School of Life Science, and the Key Laboratory of Convergence Medical Engineering System and Healthcare Technology, Ministry of Industry and Information Technology, Beijing Institute of Technology, Beijing 100081, China
General Performance Guarantee for Human Torque Estimation-Based Task-Agnostic Assistive Exoskeleton Control
Sep 30, 2026Duy Hoang, Bastien Berret, Olivier Bruneau +1Robotic Control Hip Exoskeleton Control
Universit´e Paris-Saclay, CNRS, ENS Paris-Saclay, LMF, 91190 Gif-sur-Yvette, France · Universit´e Paris-Saclay, Inria, CIAMS, 91190 Gif-sur-Yvette, France · Universit´e Paris-Saclay, ENS Paris-Saclay, LURPA, 91190 Gif-Sur-Yvette, France
ATP: Anatomical Torque with Passivity-based Control Framework for Safe Upper-Limb Exoskeleton Assistance
Aug 6, 2026Yu Chen, Gong Chen, Xiang LiAnkle Exoskeleton Control Feedback Control
Department of Automation, Tsinghua University, China · Shenzhen MileBot Robotics Co., Ltd, China

Oct 15, 2026J/K move · Enter open · S save

A Flow-Theoretic Multimodal Control Framework Fusing Mental and Physical Engagement for Upper-Limb Rehabilitation

Jun Leng, Pengcheng Li, Hanze Wang, Shuxiang Guo+1

Department of Electronic and Electrical Engineering, Southern University of Science and Technology, Shenzhen 518055, China · Aerospace Center Hospital, School of Life Science, and the Key Laboratory of Convergence Medical Engineering System and Healthcare Technology, Ministry of Industry and Information Technology, Beijing Institute of Technology, Beijing 100081, China

Assistive Robotics Continuous Control Impedance Control

Open paper AskPDF arXiv

While Assist-as-Needed (AAN) control aims to maximize user participation in robot-assisted training, existing methods relying solely on physical metrics cannot distinguish muscular inability from mental disengagement, leading to suboptimal interventions. To address this ambiguity, a Flow-theoretic multimodal AAN framework is proposed, fusing EEG-based mental engagement and sEMG-based physical engagement into an adaptive impedance control strategy. A four-quadrant fuzzy decision engine translates these dual physiological metrics into continuous control policies. This enables the upper-limb exoskeleton to inject feedforward assistance during intent-execution mismatch (e.g., motor output attenuation despite mental focus) and escalate guiding stiffness to penalize mental distraction. The framework was validated with healthy participants against a state-of-practice sEMG-only AAN baseline. Kinematically and mechanically, the proposed framework reduced robotic energy expenditure by 71.9% while maintaining precise trajectory tracking, confirming it elicited greater voluntary human effort. Physiologically, the system increased the time spent in the neuroplasticity zone (Flow State Ratio) from 4.6% to 42.8%. These results validate that decoupling and jointly regulating mental and physical states enhances the selectivity of robotic assistance, mitigating user passivity and sustaining the engagement crucial for motor learning.

Abstract

Assistive Robotics

Explore similar work

A Flow-Theoretic Multimodal Control Framework Fusing Mental and Physical Engagement for Upper-Limb Rehabilitation

General Performance Guarantee for Human Torque Estimation-Based Task-Agnostic Assistive Exoskeleton Control

ATP: Anatomical Torque with Passivity-based Control Framework for Safe Upper-Limb Exoskeleton Assistance