About

Jeremy D. Brown is the John C. Malone Associate Professor in the Department of Mechanical Engineering at Johns Hopkins University. His research explores the interface between humans and robotics, with a specific focus on medical applications and haptic feedback. Brown’s work sits at the intersection of engineering, biomechatronics, medicine, perception, and psychophysics, and involves developing novel haptic interfaces for upper-limb prostheses, minimally invasive surgical robotics, and rehabilitation robots. He leads the Haptics and Medical Robotics (HAMR) lab, where his team employs methods from human perception, motor control, neurophysiology, and biomechanics to study touch perception, especially as it relates to human-robot interaction and collaboration. His research aims to advance understanding in these areas, potentially leading to breakthroughs in rehabilitation robotics and related fields. Brown has received numerous awards, including the IEEE RAS Technical Committee on Haptics Early CAREER Award, NSF CRII and CAREER Awards, Sloan Foundation Fellowship, and the NIH Interdisciplinary Rehabilitation Engineering Career Development Program scholarship. He is a senior member of IEEE, and a member of ASME and NSBE. His work has been published in peer-reviewed journals and featured in various news outlets.

Research topics

• Computer Science
• Artificial Intelligence
• Medicine
• Physical medicine and rehabilitation
• Human–computer interaction
• Simulation
• Engineering
• Psychology
• Physical therapy
• Medical education
• Computer vision
• Cognitive psychology
• Developmental psychology
• Neuroscience
• Pedagogy

Selected publications

• Modeling and Control of a Pneumatic Soft Robotic Catheter Using Neural Koopman Operators

  ArXiv.org · 2026-03-04

  articleOpen access

  Catheter-based interventions are widely used for the diagnosis and treatment of cardiac diseases. Recently, robotic catheters have attracted attention for their ability to improve precision and stability over conventional manual approaches. However, accurate modeling and control of soft robotic catheters remain challenging due to their complex, nonlinear behavior. The Koopman operator enables lifting the original system data into a linear "lifted space", offering a data-driven framework for predictive control; however, manually chosen basis functions in the lifted space often oversimplify system behaviors and degrade control performance. To address this, we propose a neural network-enhanced Koopman operator framework that jointly learns the lifted space representation and Koopman operator in an end-to-end manner. Moreover, motivated by the need to minimize radiation exposure during X-ray fluoroscopy in cardiac ablation, we investigate open-loop control strategies using neural Koopman operators to reliably reach target poses without continuous imaging feedback. The proposed method is validated in two experimental scenarios: interactive position control and a simulated cardiac ablation task using an atrium-like cavity. Our approach achieves average errors of 2.1 +- 0.4 mm in position and 4.9 +- 0.6 degrees in orientation, outperforming not only model-based baselines but also other Koopman variants in targeting accuracy and efficiency. These results highlight the potential of the proposed framework for advancing soft robotic catheter systems and improving catheter-based interventions.

  PublisherOA PDF

• Perceiving latent dynamics: Innate and coachable visual estimation of limb damping

  bioRxiv (Cold Spring Harbor Laboratory) · 2026-03-04

  articleOpen access

  Abstract Humans are remarkably adept at extracting latent dynamic information from purely visual cues. Prior work shows that people can innately estimate differences in limb stiffness using solely their visual observation of movement, which suggests that components of mechanical impedance may be embedded within humans’ internal predictive models of movement. We tested whether humans can similarly perceive damping, a force-velocity relationship, and whether targeted coaching can enhance this visual ability. Specifically, 30 participants observed abstract two-link arm simulations with systematically varied elbow damping and rated their perceived level of damping for several trials. Participants completed two sessions separated by one of three brief coaching interventions: (1) no coaching, (2) coaching to attend to hand velocity, or (3) coaching to attend to elbow-angle velocity. Results reveal that (1) humans can innately perceive changes in arm damping using solely their visual observation of motion and (2) coaching further improved performance, with the elbow-angle coaching group showing a significantly greater increase in rating accuracy compared to the other two groups. This work extends our understanding of how action-perception coupling supports inference of mechanical impedance. Moreover, we demonstrated that perceptual strategies for estimating damping are malleable and can be systematically improved through coaching. We not only identified the visual cues observers relied on but also guided them toward more classifiable features, effectively strengthening their perceptual models of limb dynamics. Author summary Humans are remarkably adept at understanding an object’s latent dynamic properties simply by watching it move, even when the underlying forces are unseen. In this paper, we demonstrated that people can notice differences in how “damped” a moving limb is using vision alone. Moreover, we found that brief coaching helped participants focus on the most informative features, significantly improving their ability to differentiate the damping levels. These results demonstrate how people can visually infer aspects of movement that are normally thought to require physical interaction, offering insight into how the motor system links action and perception. They also show that strategies can be shaped and improved, supporting real-world healthcare applications. In stroke rehabilitation, physical therapists physically assess the resistance of a patient’s limb, so better guidance on the most relevant visual cues can help clinicians learn faster and even provide care remotely. In robot-assisted surgery, surgeons operate a console to perform procedures with limited or no force feedback, so they must estimate tissue dynamics properties largely from visual observation. Understanding how people visually estimate these dynamics can inform training for more precise surgical decisions. Overall, our findings clarify how humans interpret movement dynamics and how coaching can support more consistent and accurate perceptual decisions.

  PublisherDOI

• Modeling and Control of a Pneumatic Soft Robotic Catheter Using Neural Koopman Operators

  Open MIND · 2026-03-04

  preprint

  Catheter-based interventions are widely used for the diagnosis and treatment of cardiac diseases. Recently, robotic catheters have attracted attention for their ability to improve precision and stability over conventional manual approaches. However, accurate modeling and control of soft robotic catheters remain challenging due to their complex, nonlinear behavior. The Koopman operator enables lifting the original system data into a linear "lifted space", offering a data-driven framework for predictive control; however, manually chosen basis functions in the lifted space often oversimplify system behaviors and degrade control performance. To address this, we propose a neural network-enhanced Koopman operator framework that jointly learns the lifted space representation and Koopman operator in an end-to-end manner. Moreover, motivated by the need to minimize radiation exposure during X-ray fluoroscopy in cardiac ablation, we investigate open-loop control strategies using neural Koopman operators to reliably reach target poses without continuous imaging feedback. The proposed method is validated in two experimental scenarios: interactive position control and a simulated cardiac ablation task using an atrium-like cavity. Our approach achieves average errors of 2.1 +- 0.4 mm in position and 4.9 +- 0.6 degrees in orientation, outperforming not only model-based baselines but also other Koopman variants in targeting accuracy and efficiency. These results highlight the potential of the proposed framework for advancing soft robotic catheter systems and improving catheter-based interventions.

  DOI

• Biologically informed geometry and force distribution improve task performance in agonist/antagonist tendon-driven prosthetic hands

  medRxiv · 2026-04-06

  articleSenior author

  ABSTRACT Prosthetic devices balance functionality and usability to support activities of daily living (ADLs). However, many designs rely on rigid end effectors that, while anthropomorphic in form, lack biomimetic design principles. This mismatch increases cognitive and physical burden, reducing adoption rates. We developed the Human-inspired Actuator Modeling and Reconstruction (HAMR) process, a user-centered framework informed by individual morphology and functional needs, to generate customized agonist/antagonist tendon-actuated end effectors. Using HAMR, we created the Tendon Actuated Prosthetic Hand (TAPH), which integrates human-derived geometry with adaptive force distribution to promote natural object interaction. In a study with 12 participants without limb difference, TAPH was compared to a structurally similar tendon-actuated hand with generalized anthropomorphic geometry across three ADL tasks of varying complexity. TAPH significantly improved task performance and reduced physical effort, mental workload, and frustration, particularly during gross motor tasks. For fine motor tasks, performance improved under stable conditions but not during tasks requiring dynamic precision and continuous coordination. These findings highlight the functional benefits of biologically informed prosthesis design and support biomimetic principles in enhancing performance and user experience.

  PublisherDOI

• 3D‐Nanoprinted Fluidically Steerable Soft Robotic Microcatheters

  Advanced Robotics Research · 2026-04-28

  articleOpen access

  Endovascular interventions—i.e., minimally invasive procedures that involve navigating guidewires and/or microcatheters through a patient's vasculature to access target treatment sites—have emerged as a preferable alternative to traditional open surgery in many clinical contexts. Unfortunately, maneuverability constraints inherent to conventional guidewire‐microcatheter systems can hinder effective navigation to target sites, elevating the risk of procedural complications or failed catheterization. Soft robotic surgical instruments that harness fluidic actuation schemes hold unique promise to overcome such challenges; however, manufacturing‐induced limitations remain a critical barrier to their miniaturization, reproducibility, and clinical translation. To address these issues, in this work, a novel additive manufacturing strategy is presented to realize fluidically steerable soft robotic microcatheters via “two‐photon direct laser writing (DLW)”. As an exemplar relevant to 3‐French (1 mm‐in‐diameter) microcatheters used for transarterial chemoembolization (TACE), a 3D‐printed soft microrobotic tip is fabricated, which enables localized steering driven by microfluidic inputs while simultaneously facilitating fluidic payload delivery through a central lumen. By providing a pathway to new classes of soft robotic microsurgical instruments that enable on‐demand steerability via fluidic means, the presented strategy offers notable potential for navigating narrow, complex, tortuous, and/or delicate vasculature to enhance the safety and efficacy of endovascular therapy.

  PublisherDOI

• Lack of Disembodiment Influences Perceptual Sensitivity of Virtual Brush Illusion Paradigm

  2026-03-29

  articleSenior author

  Virtual reality (VR) has gained popularity across domains such as entertainment, training, and rehabilitation as a means of generating convincing multisensory experiences. However, the way these sensory modalities combine in VR to generate specific percepts and influence user immersion are not well understood. Illusions, particularly visuo-haptic illusions, are a valuable tool for examining how users generate internal body representations and haptic perceptions in multisensory virtual environments. Thus, we were inspired to develop a VR analog of the mirror brush illusion, where congruent and incongruent visuotactile stimuli of brushstrokes on the fingertips are presented in both the real and virtual worlds. We hypothesized that the visual stimuli would override the conflicting haptic stimuli due to participants' embodiment of their virtual hands, creating an illusory cutaneous percept. However, we found that embodiment of the virtual hands alone is insufficient to generate this illusory haptic percept. Instead, it appears that strong embodiment of the virtual hand, along with strong disembodiment of the real hand, may be necessary to induce the illusion. These findings present novel evidence that visual dominance in humans does not always outcompete haptic sensation, and suggest methods by which immersion in virtual reality can be strengthened by specific multisensory input.

  PublisherDOI

• Teleoperator Coupling Dynamics Impact Human Motor Control Across Pursuit Tracking Speeds

  IEEE Transactions on Haptics · 2025-01-01

  articleSenior author

  Robotic teleoperators introduce novel electromechanical dynamics between the user and the environment. While considerable effort has focused on minimizing these dynamics, we lack a robust understanding of their impact on user task performance across the range of human motor control ability. Here, we utilize a 1-DoF teleoperator testbed with interchangeable mechanical and electromechanical couplings between the leader and follower to investigate to what extent, if any, the dynamics of the teleoperator influence performance in a visual-motor pursuit tracking task. We recruited N = 30 participants to perform the task at frequencies ranging from 0.55-2.35 Hz, with the testbed configured into Mechanical, Unilateral, and Bilateral configurations. Results demonstrate that tracking performance at the follower was similar across configurations. However, participants' adjustment at the leader differed between Mechanical, Unilateral, and Bilateral configurations. In addition, participants applied different grip forces between the Mechanical and Unilateral configurations. Finally, participants' ability to compensate for coupling dynamics diminished significantly as execution speed increased. Overall, these findings support the argument that humans are capable of incorporating teleoperator dynamics into their motor control scheme and producing compensatory control strategies to account for these dynamics; however, this compensation is significantly affected by the leader-follower coupling dynamics and the speed of task execution.

  PublisherDOI

• Understanding the Utility of State-Based Haptic Feedback in Tendon-Driven Anthropomorphic Prostheses

  IEEE Transactions on Neural Systems and Rehabilitation Engineering · 2025-01-01 · 1 citations

  articleOpen accessSenior author

  Haptic feedback has demonstrated utility in traditional prosthetic devices, however, it is unclear to what extent haptic feedback improves functionality in an anthropomorphic agonist/antagonist tendon-actuated design. We investigate the impact of state-based haptic feedback in an agonist/antagonist tendon-driven anthropomorphic prosthesis by proportionally mapping haptic sensations of the tension in the tendons during actuation. N =24 participants without limb loss were recruited to perform a grasp and transfer task using a mock prosthesis across three conditions: no haptic feedback, skin stretch feedback, and vibrotactile feedback. We hypothesized that haptic feedback of tendon tension would improve task performance and that skin-stretch feedback would outperform the vibrotactile condition due to the modality-matched similarities of tension and stretch. Results highlight that vibrotactile feedback resulted in significantly more object transfers than skin stretch feedback or no feedback. However, skin stretch feedback had a significantly higher transfer efficiency than vibrotactile feedback, demonstrating that different haptic modalities uniquely affect task performance. This study is the first to demonstrate that feedback of tendon tension in a tendon-driven prosthesis has significant utility and improves task performance establishing a need for further exploration of haptic integration in tendon-actuated systems.

  PublisherDOI

• HaptiKart: An engaging videogame reveals elevated proprioceptive bias in individuals with autism spectrum disorder

  PLOS Digital Health · 2025-06-18 · 1 citations

  articleOpen access

  An overreliance on proprioceptive (intrinsic) sensory input from the body, compared to visual (extrinsic) input from the environment, may underpin core features of autism spectrum disorder (ASD). We developed an engaging videogame ("HaptiKart") as a tool to examine differences in sensory-motor bias (proprioceptive vs. visual) in children and adults with ASD and whether bias correlates with age, core autism features, and intellectual ability. Eighty-one participants (33 ASD, 48 typically-developing, TD) aged 8-31 years played "HaptiKart," a driving videogame with a force-feedback steering wheel that provided "steering assist" during gameplay. In separate trials, proprioceptive and visual feedback were selectively delayed, and differences in driving error between the conditions were used to calculate perceptual bias scores. Effects of autism diagnosis and age on bias scores were examined, controlling for sex, as were associations of perceptual bias with autism symptom severity (ADOS-2, SRS-2), attention-deficit symptom severity (Conners4 ADHD Total Scores) ratings, and IQ (general ability index, GAI). The ASD group exhibited significantly higher proprioceptive bias than did the TD group (p = 0.002). There was a trend for decreasing proprioceptive bias with age, but no significant diagnosis-by-age interaction. Increased proprioceptive bias correlated with higher autism severity and with lower IQ, but not ADHD symptoms. HaptiKart provides a highly scalable approach for measuring sensory-motor bias, revealing that individuals with ASD show elevated proprioceptive bias, correlating with autism severity. HaptiKart's sensory-motor bias measure may thereby serve as a digital biomarker for addressing autism heterogeneity in ways that can improve targeted intervention.

  PublisherDOI

• Autonomous Closed-Loop Control for Robotic Soft Tissue Electrosurgery Using RGB-D Image Guidance

  IEEE Transactions on Medical Robotics and Bionics · 2025-06-25 · 1 citations

  articleOpen access

  Oral cavity cancer, a common head and neck cancer, is typically treated through precise tumor excision via electrosurgery. Autonomous robotic electrosurgery has demonstrated the potential to achieve more accurate and consistent resection margins compared to manual methods, thereby improving surgical outcomes. However, current autonomous systems face challenges in tracking tissue deformation during electrosurgical cutting due to unpredictable and complex soft tissue dynamics. Failure to monitor and adapt to tissue deformation can significantly compromise resection precision. This paper presents an autonomous closed-loop robotic electrosurgery system to enhance surgical precision via 3D tissue tracking and image-based feedback control utilizing a Red Green Blue – Depth (RGB-D) sensor. The developed 3D tissue tracker employs CoTracker, a deep learning-based model for markerless tracking, complemented by a tool-occlusion algorithm to achieve tissue deformation tracking with no prior knowledge of the tissue model. The estimated deformation is fed into a fuzzy logic controller, which dynamically adjusts the cutting velocity to minimize cutting error during electrosurgery. The system’s efficacy was validated using ex vivo porcine tongues, demonstrating a 55% reduction in average cutting error (from 1.2 mm to 0.54 mm, p<0.001) in closed-loop operations (N=6) compared to open-loop cutting without feedback control (N=3). The results demonstrate the effectiveness of image-based closed-loop control in improving margin accuracy, a key factor in reducing the likelihood of cancer recurrence.

  PublisherOA PDFDOI

Recent grants

• CRII: CHS: Improving Dexterous Manipulation with Telerobots Through Operator-Sensitive Haptic Display

  NSF · $223k · 2017–2021

• CHS: Small: Understanding Environment Perception and Task Performance in Human-in-the-Loop Tele-robotic Systems (HiLTS)

  NSF · $545k · 2019–2025

Frequent coauthors

• Katherine J. Kuchenbecker

  Max Planck Institute for Intelligent Systems

  109 shared

• Yan Wu

  100 shared

• Knut Drewing

  100 shared

• William Frier

  Imperial College London

  100 shared

• Albert Star

  Université Paris Cité

  100 shared

• Rebecca Friesen

  Scuola Superiore Sant'Anna

  100 shared

• Mehdi Ammi

  Laboratoire Cognitions Humaine et Artificielle

  100 shared

• Seungmoon Choi

  100 shared

Labs

• Haptics and Medical Robotics (HAMR) LaboratoryPI

Education

• Doctor of Philosophy, Mechanical Engineering

  University of Michigan

  2014

• Master of Science, Mechanical Engineering

  University of Michigan

  2012

• Bachelor of Science, Mechancial Engineering (Dual Degree Engineering Program)

  University of Michigan

  2008

• Bachelor of Science, Applied Physics/Dual Degree Engineering

  Morehouse College

  2006

Awards & honors

• IEEE RAS Technical Committee on Haptics Early CAREER Award
• NSF CRII Award
• NSF CAREER Award
• Sloan Foundation Fellowship
• Penn Postdoctoral Fellowship for Academic Diversity