cs.ROMar 13, 2026

Skill-informed Data-driven Haptic Nudges for High-dimensional Human Motor Learning

Authors: Ankur Kamboj, Rajiv Ranganathan, Xiaobo Tan, Vaibhav Srivastava

Organizations: Department of Electrical and Computer Engineering, Michigan State University, USA. · Department of Kinesiology and the Department of Mechanical Engineering, Michigan State University, USA.

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

In this work, we propose a data-driven framework to design optimal haptic nudge feedback leveraging the learner's estimated skill to address the challenge of learning a novel motor task in a high-dimensional, redundant motor space. A nudge is a series of vibrotactile feedback delivered to the learner to encourage motor movements that aid in task completion. We first model the stochastic dynamics of human motor learning under haptic nudges using an Input-Output Hidden Markov Model (IOHMM), which explicitly decouples latent skill evolution from observable performance measures. Leveraging this predictive model, we formulate the haptic nudge feedback design problem as a Partially Observable Markov Decision Process (POMDP). This allows us to derive an optimal nudging policy that minimizes long-term performance cost and implicitly guides the learner toward superior skill states. We validate our approach through a human participant study (N=30) involving a high-dimensional motor task rendered through a hand exoskeleton. Results demonstrate that participants trained with the POMDP-derived policy exhibit significantly accelerated movement efficiency and endpoint accuracy compared to groups receiving heuristic-based feedback or no feedback. Furthermore, synergy analysis reveals that the POMDP group discovers efficient low-dimensional motor representations more rapidly.

Figures & tables

Figure 1: Target Capture Task: (a) shows a participant performing our target capture task with the SenseGlove DK1 exoskeleton strapped to their right hand. (b) and (c) show the cursor trajectories during the first and eighth blocks of the game with the red dots representing the target locations. The trajectories get straighter as the participant learns to control the cursor by the end of the experiment session. (d) shows how the performance metrics, Reaching Error and Straightness of Trajectory, are computed.
Figure 2: Simplified IOHMM: Human motor learning behavior modeled as an IOHMM shows how the hidden motor skill state hkh_{k} transitions to hk+1h_{k+1} under the effect of the input xk={uk:slopek,ak:nudgek}\bf{x}_{k}=\{u_{k}:\texttt{slope}_{k},a_{k}:\texttt{nudge}_{k}\} , leading to the emissions ok={REk,SoTk}o_{k}=\{\texttt{RE}_{k},\texttt{SoT}_{k}\} under the influence of the same input xk\bf{x}_{k} .
Figure 3: Estimated Behavioral Model: (a) Predicted RE and SoT means out of the emission model for the ordered skill states, and (b) State Transition Matrices extracted from the estimated human motor learning behavioral model show how participants tend to transition from low latent skill states to high latent skill states under the effect of the input slope angles uk=−1.57,−1.11u_{k}=-1.57,-1.11 (corresponding to target pairs {0,2}\{0,2\} ) and the nudged finger indices ak=0,⋯ ,5a_{k}={0,\cdots,5} . The latent skill states are ordered best-to-worst (0−6)(0-6) based on their mean output RE emission values. The nudge=0\texttt{nudge}=0 value represents no nudging.
Figure 4: Movement Efficiency: (a) Mean SoT curves (with 95%95\% confidence intervals) of the three group participants, and (b) the mean number of trials (along with the scatter plot taken across participants) to converge to a specific SoT threshold under the three nudging policies shows fastest convergence for participants trained on POMDP-based optimal nudging policy, followed by the heuristic policy. The low p-values from two-tailed tests from the linear mixed model fits show a decrease in SoT across trials and in the number of trials required by the POMDP group for various SoT thresholds. Each scatter plot point represents the number of trials taken by a participant to achieve the SoT threshold.
MetricsPOMDP vs ControlPOMDP vs Heuristic
pbonfp_{\textup{bonf}}dcohend_{\textup{cohen}}pbonfp_{\textup{bonf}}dcohend_{\textup{cohen}}
SoT Curve0.1053-0.53680.5002-0.6048
 0.340.5290-0.51511.0000-0.1277
SoT0.390.2865-0.68451.0000-0.5148
Thresholds0.440.0875-0.96981.0000-0.2950
 0.490.0796-0.99121.0000-0.1619
Table 1: Two-tailed Contrasts for SoT Group Differences. The p-values for the hypothesis that the POMDP group has different SoT values across gameplay trials and different number of trials to achieve thresholds compared to the control and the heuristic group are reported. The results show that the POMDP group achieves lower SoT values on average, and takes statistically significantly (at significance level α=0.10\alpha=0.10 ) fewer gameplay trials to achieve SoT thresholds early in the training.
Figure 5: Endpoint Accuracy: (a) Mean RE curves (with 95%95\% confidence intervals) of the three group participants, and (b) the mean number of trials (along with the scatter plot taken across participants) to converge to different RE thresholds under the three nudging policies shows fastest convergence for participants trained on POMDP-based optimal nudging policy. The low two-tailed test p-values from the linear mixed model fits show a decrease in RE across trials and in the number of trials required by the POMDP group for all RE thresholds. Each scatter plot point represents the number of trials taken by a participant to achieve that RE threshold.
Figure 6: Synergy Analysis: (a) Variance accounted for (VAF) by the first four principal components (PC), and (b) the number of PCs required to explain more than 90%90\% variance in joint data across the three participant groups reveal that the POMDP group achieved a higher VAF by top-PCs (especially in later phases) and required fewer PCs to explain >90%>90\% variance in the data on average, signifying a more rapid discovery of the task relevant subspace towards expedited skill acquisition. Training blocks are consolidated into phases, with each phase consisting of two consecutive blocks.
MetricsPOMDP vs ControlPOMDP vs Heuristic
pbonfp_{\textup{bonf}}dcohend_{\textup{cohen}}pbonfp_{\textup{bonf}}dcohend_{\textup{cohen}}
RE Curve0.0642-0.38630.0249-0.5899
 0.141.0000-0.30251.0000-0.0284
RE0.191.0000-0.04001.0000-0.0153
Thresholds0.340.1705-0.81440.1223-0.8932
 0.390.3765-0.61160.4802-0.5433
Table 2: Two-tailed Contrasts for RE Group Differences. The p-values for the hypothesis that the POMDP group has different RE values across gameplay trials and different number of trials to achieve thresholds compared to the control and the heuristic group are reported. The results show that the POMDP group achieves statistically significantly (at significance level α=0.10\alpha=0.10 ) lower RE values on average, and takes fewer gameplay trials to achieve RE thresholds early in the training.
Figure 7: Learning Dynamics and Convergence Analysis: Expected latent state evolution across trials and the number of trials to reach a steady state mastery level ( E[sk]<3\mathbb{E}[s_{k}]<3 ) reveals expedited convergence of POMDP group to better performing states and a statistically significantly lower number of trials to mastery, compared to heuristic and control groups.
More in this topic

Haptic Feedback

Haptic feedback conveys touch, force, motion, or material properties through devices that let people and robots perceive and respond to interaction.

See all 60 papers

Explore similar work

CardsList
Aug 4, 2026

AIDE: Automated Instruction via Distilled Expertise for Reference-Free Motor Skill Coaching

Generating natural-language coaching feedback on motor skills can accelerate learning, yet expert coaches are scarce and expensive. Existing reference-based methods require expert demonstrations at both training and inference time, limiting practical deployment. We propose AIDE (Automated Instruction via Distilled Expertise), a framework that exploits expert references only during training and generates feedback from a learner's pose sequence alone at inference. A teacher model first learns to generate feedback from paired learner-expert poses via a frozen language model, producing separate learner tokens and difference tokens that encode the learner-expert difference. A student model then inherits the teacher's encoder and weight initialization, replacing the explicit expert comparison with an auxiliary module that produces complementary tokens from the learner's pose alone. On the ExpertAF dataset, AIDE outperforms reference-free baselines on most metrics and performs comparably to methods requiring expert demonstrations at both training and inference, with LLM-based evaluation supporting these findings.

Yoshiki Ito

Research & Development Group Hitachi, Ltd. Tokyo, Japan

Text Generation
AskPDF
Aug 17, 2026

Lambda-Hold Control: Human-Like Movement Emerges from a Minimal Task Reward in Predictive Musculoskeletal Simulation

The massive overactuation in the human musculoskeletal system makes it challenging to train musculoskeletal models to generate human-like motion via reinforcement learning, primarily because exploration in the resulting high-dimensional and redundant action space is extremely inefficient. To address this problem, we propose the λλ-hold controller, inspired by the equilibrium-point (EP) hypothesis, which has been widely supported by extensive evidence from human motor control studies. The policy's control variable is the per-muscle EP threshold length λλ, from which a stretch-reflex recruitment law computes the muscle excitations automatically. Holding each λλ over an interval of the gait phase also sharply reduces the frequency at which the policy must be queried. Consequently, the controller, to our knowledge for the first time, enables a muscle-actuated skeletal model to learn human-like sprinting using only a minimal reward within an hour of training. The efficient exploration through the proposed λλ-hold controller is not merely an engineering trick but an approach grounded in physiology, bringing together the EP hypothesis, intermittent control, and optimal feedback control. Beyond encapsulating human-like behavior in predictive simulation, this achievement contributes to developing a learnable model of the human motor controller.

Jun Hyuk Lee, Chihyeong Lee, Jooeun Ahn

Department of Physical Education, Seoul National University, Seoul, Republic of Korea · Institute of Sport Science, Seoul National University, Seoul, Republic of Korea · SNU Robotics Institute, Seoul National University, Seoul, Republic of Korea

RL for Legged LocomotionMusculoskeletal ModelingFeedback Control
AskPDF
May 23, 2026

MuGen: Multi-Skill Generative Locomotion Controller for Humanoid Robots

This paper presents MuGen, a data-driven framework for learning and deploying multi-skill locomotion on humanoid robots. MuGen enables a robot to perform expressive motions like humans under the guidance of example motion sequences. To achieve this, we employ vector-quantized autoencoders (VQ-VAEs) trained with model-based reinforcement learning, resulting in a generative representation of locomotion that captures key patterns of human motion from hours of heterogeneous human performance data. We employ a teacher-student learning framework and develop a new policy distillation strategy to enable a deployable student policy learning this efficient latent representation. This policy allows the robot to track and mimic unseen human motions and further enables the robot to reuse the learned latent space for other tasks. We demonstrate the effectiveness of our framework through a diverse set of motions and accurate execution.

Yusen Feng, Xiang Wang, Heyuan Yao +7

Peking University

Latent Action LearningHumanoid Robot LocomotionRobot Skill LearningRobot Motion GenerationRobot Imitation Learning
AskPDF