Therapist-Robot-Patient Physical Interaction is Worth a Thousand Words: Enabling Intuitive Therapist Guidance via Remote Haptic Control

Beatrice Luciani; Alex van den Berg; Matti Lang; Alexandre L. Ratschat; Laura Marchal-Crespo

cs.ROFeb 25, 2026

Therapist-Robot-Patient Physical Interaction is Worth a Thousand Words: Enabling Intuitive Therapist Guidance via Remote Haptic Control

Authors: Beatrice Luciani, Alex van den Berg, Matti Lang, Alexandre L. Ratschat, Laura Marchal-Crespo

Organizations: Department of Cognitive Robotics, Delft University of Technology, Delft, The Netherlands · Department of Rehabilitation Medicine, Erasmus Medical Center, Rotterdam, The Netherlands.

Abstract

Robotic systems can enhance the amount and repeatability of physically guided motor training. Yet their real-world adoption is limited, partly due to non-intuitive trainer/therapist-trainee/patient interactions. To address this gap, we present a haptic teleoperation system for trainers to remotely guide and monitor the movements of a trainee wearing an arm exoskeleton. The trainer can physically interact with the exoskeleton through a commercial handheld haptic device via virtual contact points at the exoskeleton's elbow and wrist, allowing intuitive guidance. Thirty-two participants tested the system in a trainer-trainee paradigm, comparing our haptic demonstration system with conventional visual demonstration in guiding trainees in executing arm poses. Quantitative analyses showed that haptic demonstration significantly reduced movement completion time and improved smoothness, while speech analysis using large language models for automated transcription and categorization of verbal commands revealed fewer verbal instructions. The haptic demonstration did not result in higher reported mental and physical effort by trainers compared to the visual demonstration, while trainers reported greater competence and trainees lower physical demand. These findings support the feasibility of our proposed interface for effective remote human-robot physical interaction. Future work should assess its usability and efficacy for clinical populations in restoring clinicians' sense of agency during robot-assisted therapy.

Figures & tables

Figure 1: Overview of our teleoperation therapist-in-the-loop haptic system. The Trainee (right) wears the ARMin exoskeleton, which is remotely controlled by the Trainer (left) using a Sigma.7 haptic device. The Trainer wears a headset and is immersed in VR, where the digital twin of the exoskeleton and the Trainee’s avatar are visualized. The proposed haptic framework is evaluated in the Haptic Demonstration (HD) condition and compared with a conventional Visual Demonstration (VD) approach. In the VD condition, the Trainer is provided with a webcam, and the Trainee visualizes the Trainer’s movements on a screen in real time. The system includes three interconnected computers, represented in the figure through three rounded rectangular blocks labeled “Host computer”, “Linux computer”, and “Target (ARMin) computer”. In the image,

\vec{X_{s}}

denotes the Cartesian position of the Sigma.7 end-effector,

\vec{F_{s,a}}

is the interaction force computed on the Sigma.7 side by the Teleoperation Controller,

\vec{F_{a}}

the force computed on the ARMin side, and

J_{1},\dots,J_{N}

are the ARMin joint positions recorded in real-time.

Figure 2: (a) Remote visualization of the virtual environment displayed to the Trainer. The visualization includes the ARMin digital twin [ 34 ] and a representation of the Trainee’s arm and torso (male or female). The red spheres on the avatar’s wrist and elbow joints represent the graspable points, and the yellow cube is the representation of the end-effector of the Sigma.7 manipulated by the Trainer. (b) Snapshot of the five target arm poses that the Trainee was guided to execute during the experiment.

	Estimate	Std. Error	95% CI	df	t value	Pr( $>\|t\|$ )	$p_{FDR}$
(Intercept)	20.81	1.90	[16.87, 24.75]	22.35	11.00	1.72e-10	1.03e-09***
HD	-3.84	1.46	[-6.71, -0.97]	439.29	-2.63	0.009	0.018*
Block	-0.70	0.66	[-2.00, 0.60]	439.10	-1.07	0.287	0.430
HD_first	0.65	2.50	[-4.63, 5.93]	16.97	0.26	0.798	0.798
HD:HD_first	4.88	1.53	[1.87, 7.89]	439.17	3.19	0.002	0.005**
HD:Block	-0.62	0.95	[-2.49, 1.25]	439.24	-0.66	0.509	0.611

Table 1: Results from the LMM with Completion Time. CI: Wald 95% confidence intervals. Condition Order:2 when the HD condition is the first. For clarity, in the table, we reported “HD_first” instead of Condition Order. .(

p<0.1

), *(

p<0.05

), **(

p<0.01

), ***(

p<0.001

	Elbow SPARC [-]							Wrist SPARC [-]
	Est.	Std. Er.	95% CI	df	t value	Pr( $>\|t\|$ )	$p_{\mathrm{FDR}}$	Est.	Std. Er.	95% CI	df	t value	Pr( $>\|t\|$ )	$p_{\mathrm{FDR}}$
(Intercept)	-4.20	0.34	[-4.87, -3.53]	28.01	-12.36	7.40e-13	4.44e-12***	-3.92	0.33	[-4.57, -3.27]	25.08	-11.78	1.02e-11	6.11e-11***
HD	0.70	0.31	[0.09, 1.31]	450.15	2.28	0.023	0.046*	0.54	0.30	[-0.05, 1.13]	447.10	1.95	0.052	0.104
HD_first	0.05	0.41	[-0.75, 0.85]	19.43	0.11	0.912	0.966	0.19	0.41	[-0.61, 1.00]	18.31	0.45	0.654	0.669
Block	0.25	0.14	[-0.02, 0.52]	450.16	1.80	0.073	0.110	0.15	0.12	[-0.09, 0.39]	447.19	1.19	0.235	0.353
HD:HD_first	-0.97	0.32	[-1.60, -0.34]	450.11	-3.04	0.003	0.007*	-1.04	0.29	[-1.61, -0.47]	447.08	-3.61	3.41e-04	0.001**

Table 2: Results of the LMMs for smoothness —evaluated through SPARC— at elbow and wrist graspable points. CI: Wald 95% confidence intervals. Condition Order:2 when the HD condition is the first. For clarity, in the table, we reported “HD_first” instead of Condition Order. .(

p<0.1

), *(

p<0.05

), **(

p<0.01

), ***(

p<0.001

Figure 3: Summary of the results of the speech analysis, with numbers of occurrences for Instruction, Feedback, and Null under Haptic (blue) and Visual (red) conditions. The boxes show the medians and interquartile ranges.

Total Instructions
	Estimate	Std. Error	95% CI	df	t value	Pr( $>\|t\|$ )
(Intercept)	32.81	3.59	[25.69, 39.94]	99.89	9.14	7.67e-15 ***
HD	-3.22	2.37	[-7.93, 1.48]	87.15	-1.36	0.177
Block	-4.47	1.45	[-7.37, -1.58]	87.23	-3.08	0.003 **
Percentage Instructions
(Intercept)	53.59	3.94	[45.77, 61.40]	103.66	13.61	$<$ 2e-16 ***

Table 3: Results of the LLMs for Total Instructions and Percentage of Instructions provided by Trainers to Trainees. CI: Wald 95% confidence intervals. *(

p<0.05

), **(

p<0.01

), ***(

p<0.001

Figure 4: Summary of the RTLX scores, divided per dimension and role within the dyad. The boxes show the medians and interquartile ranges. Significant differences, based on the results of the LLMs, are reported: *(p < 0.05), **(p < 0.01), ***(p < 0.001).

Figure 5: Average Effort and Perceived Competence subscales of the IMI. Significant differences, based on the results of the LLMs, are reported: *(p < 0.05), **(p < 0.01), ***(p < 0.001).

Explore similar work

Semantic Haptic Feedback Enhances Dexterous Robotic Teleoperation
Aug 3, 2026Bingjian Huang, Sahar Aseeri, Jonas Schmidtler +9Robot Teleoperation Haptic Feedback
Learning Varying Physical Therapist-Patient Interactions for Robot-mediated Upper Limb Task-Specific Training
Aug 17, 2026Jia Quan Loh, Vincent Crocher, Marlena Klaic +2Assistive Robotics Human-Robot Interaction
Human Robotics Laboratory, Department of Mechanical Engineering, The University of Melbourne · Melbourne School of Health Sciences, The University of Melbourne
HaptiNet: Networked Haptic Robots Enable Physical Co-presence in Geographically-Unconstrained Rehabilitation
Sep 4, 2026Chenyang Sun, Mingjie Dong, Haodong Deng +18Assistive Robotics Healthcare
Department of Biomedical Engineering, Southern University of Science and Technology, Shenzhen, 518055, China. · School of Biomedical Engineering, Harbin Institute of Technology, Shenzhen, 518055, China. · Beijing Key Laboratory of Advanced-Manufacturing Technology, College of Mechanical-&Energy Engineering, Beijing University of Technology, Beijing, 100124, China. +9

Aug 3, 2026J/K move · Enter open · S save

Semantic Haptic Feedback Enhances Dexterous Robotic Teleoperation

Bingjian Huang, Sahar Aseeri, Jonas Schmidtler, Joseph Zhang+8

Robot Teleoperation Haptic Feedback Robotic Manipulation

Open paper AskPDF arXiv

In robot teleoperation, haptic feedback can be used to help human operators accomplish dexterous manipulation tasks. However, existing haptic feedback methods try to replicate high-fidelity sensory haptics that are felt in real world interactions, which are constrained by the sensing and feedback hardware capability and may lead to higher workload. To addresses these limitations, this work introduces semantic haptics for teleoperation, which uses abstract haptic patterns to convey critical information about robot states. We categorize robot states into "Confirmations" and "Exceptions", implement a modular haptic rendering pipeline in robot simulation, and deliver semantic haptic feedback to operators through pneumatic and vibrotactile wristbands. This simplifies hardware requirements and enables one-to-many mappings between haptic patterns and robot states. Through three evaluation studies, we identify the most effective semantic haptic design for a common pick and place teleoperation task and compare semantic haptics to other teleoperation feedback approaches including sensory haptics and visual feedback. Results suggest that while semantic haptics performs similarly as other feedback in unimanual tasks, it achieves superior performance in bimanual tasks, with reduced task workload, increased situational awareness, and overall preference.

Abstract

Explore similar work

Semantic Haptic Feedback Enhances Dexterous Robotic Teleoperation

Learning Varying Physical Therapist-Patient Interactions for Robot-mediated Upper Limb Task-Specific Training

HaptiNet: Networked Haptic Robots Enable Physical Co-presence in Geographically-Unconstrained Rehabilitation