About

Robert Gregg is a Professor of Mechanical Engineering, with additional titles as Professor of Robotics and Professor of Electrical and Computer Engineering at the University of Michigan. He serves as the Associate Director for Graduate Education at Michigan Robotics and is the Director of the Locomotor Control Systems Laboratory. His research interests include wearable robots, legged robots, prosthetics and orthotics, bipedal locomotion, nonlinear control theory, and rehabilitation engineering. Gregg's academic background includes a PhD and MS from the University of Illinois at Urbana-Champaign, earned in 2010 and 2007 respectively, and a BS from the University of California – Berkeley obtained in 2006. His work focuses on advancing technologies related to mobility, control systems, and rehabilitation, contributing to the development of innovative robotic systems and control methodologies.

Research topics

• Computer Science
• Biology
• Physics
• Surgery
• Anatomy
• Classical mechanics
• Medicine

Selected publications

• Experiment-free learning of exoskeleton assistance remains an unsolved problem

  bioRxiv (Cold Spring Harbor Laboratory) · 2026-04-06

  article

  Abstract In "Experiment-free exoskeleton assistance via learning in simulation", Luo et al. [1] present an ambitious framework for developing exoskeleton controllers through reinforcement learning exclusively in computer simulation. The authors report that a control policy trained on a small dataset from one subject was directly transferred to physical hardware, reducing human metabolic cost during walking, running, and stair climbing by more than any prior device. If confirmed, this would represent a major breakthrough for the field of wearable robotics and their clinical applications. However, a close examination of the published materials casts doubt on these claims. The reported experimental results violate physiological limits on the relationship between mechanical power and muscle energy use during gait 2,3,4 . The algorithmic claims are surprising and cannot be verified; in contrast with established replicability standards in machine learning 5,6 , executable code has not been made available. We conclude that the goals of this study have not yet been verifiably achieved and make recommendations for avoiding publication errors of this type in the future.

  PublisherDOI

• Task-Agnostic Exoskeleton Control Supports Elderly Joint Energetics during Hip-Intensive Tasks

  arXiv (Cornell University) · 2026-03-23

  preprintOpen accessSenior author

  Age-related mobility decline is frequently accompanied by a redistribution of joint kinetics, where older adults compensate for reduced ankle function by increasing demand on the hip. Paradoxically, this compensatory shift typically coincides with age-related reductions in maximal hip power. Although robotic exoskeletons can provide immediate energetic benefits, conventional control strategies have limited previous studies in this population to specific tasks such as steady-state walking, which do not fully reflect mobility demands in the home and community. Here, we implement a task-agnostic hip exoskeleton controller that is inherently sensitive to joint power and validate its efficacy in eight older adults. Across a battery of hip-intensive activities that included level walking, ramp ascent, stair climbing, and sit-to-stand transitions, the exoskeleton matched biological power profiles with high accuracy (mean cosine similarity 0.89). Assistance significantly reduced sagittal plane biological positive work by 24.7% at the hip and by 9.3% for the lower limb, while simultaneously augmenting peak total (biological + exoskeleton) hip power and reducing peak biological hip power. These results suggest that hip exoskeletons can potentially enhance endurance through biological work reduction, and increase functional reserve through total power augmentation, serving as a promising biomechanical intervention to support older adults' mobility.

  PublisherDOI

• Task-Agnostic Exoskeleton Control Supports Elderly Joint Energetics during Hip-Intensive Tasks

  arXiv (Cornell University) · 2026-03-23

  articleOpen accessSenior author

  Age-related mobility decline is frequently accompanied by a redistribution of joint kinetics, where older adults compensate for reduced ankle function by increasing demand on the hip. Paradoxically, this compensatory shift typically coincides with age-related reductions in maximal hip power. Although robotic exoskeletons can provide immediate energetic benefits, conventional control strategies have limited previous studies in this population to specific tasks such as steady-state walking, which do not fully reflect mobility demands in the home and community. Here, we implement a task-agnostic hip exoskeleton controller that is inherently sensitive to joint power and validate its efficacy in eight older adults. Across a battery of hip-intensive activities that included level walking, ramp ascent, stair climbing, and sit-to-stand transitions, the exoskeleton matched biological power profiles with high accuracy (mean cosine similarity 0.89). Assistance significantly reduced sagittal plane biological positive work by 24.7% at the hip and by 9.3% for the lower limb, while simultaneously augmenting peak total (biological + exoskeleton) hip power and reducing peak biological hip power. These results suggest that hip exoskeletons can potentially enhance endurance through biological work reduction, and increase functional reserve through total power augmentation, serving as a promising biomechanical intervention to support older adults' mobility.

  PublisherOA PDF

• Controlling Powered Prosthesis Joint Impedance Over Continuous Stance Transitions Between Walking and Stair Ascent/Descent

  IEEE Transactions on Neural Systems and Rehabilitation Engineering · 2026-01-01

  articleOpen accessSenior author

  Transition strides between level-ground walking and stairs are important parts of everyday locomotion that help maintain balance while sustaining the momentum of the user between these activities of daily living. However, most individuals with transfemoral amputations are unable to perform these continuous transitions with their conventional passive prostheses, instead being forced to initiate transitions with a specific leg or pause at the threshold of the staircase. Powered prostheses have the potential to allow for continuous transitions due to their ability to provide positive work and active control during level-ground walking and stair locomotion, but modern impedance control approaches switch discretely between steady-state controllers instead of emulating continuous joint biomechanics. This work presents a phase-based hybrid kinematic impedance controller that provides biologically-inspired knee and ankle impedance during continuous stance-phase transitions between level-ground walking and stair ascent/descent, assuming high-level knowledge of the transition stride. In an offline analysis of N=12 participants, we show that our continuous stance transition modeling approach significantly outperforms a typical discrete switching strategy in most cases. To experimentally implement the transition model, we define a common thigh-based phase variable for both steady-state and transition strides, giving the user control over prosthesis stride progression. Pilot experiments with two K4 transfemoral amputee participants using a powered knee-ankle prosthesis demonstrate biomimetic kinematic/kinetic features during stair ascent/descent transitions for two stair incline configurations, without subject-specific tuning of control parameters.

  PublisherDOI

• Optimal Energy Shaping and Force Amplification Framework for Task-Agnostic, Biomimetic Ankle Exoskeletons

  2026-02-10

  articleOpen accessSenior author

  Task-agnostic controllers for partial-assist lower-limb exoskeletons aim to reliably mimic biological torque while seamlessly adapting to changing movement patterns. However, current approaches relying on hidden state estimators or neural networks lack explainability and safety guarantees, while force amplification methods risk instability with an inherent trade-off between sensitivity and robustness to control inputs. Energy shaping control uses a kinematic model-based framework to provide predictable, stable assistance, though its traditional passive form limits biomimetic performance. Previous work relaxed the strict passivity requirements to improve biomimicry but reduced the stability guarantees. This paper presents an optimization-based extension of the energy-shaping control framework that combines the stability benefits of energy shaping with the intuitive biomimicry of force amplification. Our framework enables controlled trade-offs between sensitivity to changing human impedance and high performance through adjustable cost contributions of force amplification and model-based terms. We provide theoretical guarantees of closed-loop stability to an invariant set under human joint impedance control, supported by empirical validation of stability characteristics of an ankle exoskeleton under varying controller passivity constraints. A study of ten able-bodied participants using bilateral ankle exoskeletons demonstrates that the biomimetic controller reduced biological ankle torque by 19.1% across various activities of daily life.

  PublisherDOI

• Effects of a Powered Knee–Ankle Prosthesis on Intact Joint Biomechanics Across Sustained Activities of Daily Life: A Case Series

  IEEE Transactions on Neural Systems and Rehabilitation Engineering · 2026-01-01

  articleOpen accessSenior author

  Lower-limb prosthesis users often overuse their intact joints due to the lack of positive work generated by their devices. This overreliance has been shown to increase joint loading, degeneration, and pain. While powered prostheses can generate positive work and therefore reduce this burden, clinical studies of commercialized single-joint devices have demonstrated inconsistent results. Recently, prototype powered knee and ankle prostheses have shown more consistent advantages over passive devices in laboratory settings. Most of the studies, however, focus on the biomechanics of the prosthesis rather than its impact on the user's joints, study isolated activities, and/or do not replicate the demands of continuous real-world use. This case series analyzes the intact joint moments and work for N=3 above-knee amputee subjects using a powered knee-ankle prosthesis vs. their prescribed passive device during a continuous, sustained sequence of the primary activities of daily life. The powered prosthesis decreased peak hip flexion moment (but increased peak extension moment) during level walking, and decreased peak knee extension moment for all other activities. For at least two of the three subjects, the powered prosthesis decreased total positive work across the intact joints during ascent activities (stair ascent, sit-to-stand) and decreased negative total work for descent activities (stair descent, stand-to-sit). This case series suggests that powered knee-ankle prostheses have the potential to reduce overuse of intact joints in emulated real-world conditions.

  PublisherDOI

• Human-Interface Dynamics of Knee Exoskeletons with Lateral and Anteroposterior Attachment

  2025-05-12 · 1 citations

  articleOpen access

  Assistive lower-body exoskeletons aim to improve quality of life for broad populations including older adults and people in physically exhausting manual jobs. By applying torque to augment human motion with backdrivable actuators, these devices can reduce human joint effort without restricting volitional motion. However, these backdrivable actuators are coupled by mechanical interfaces to soft tissues of the human body that together introduce resonator dynamics that can delay or diminish the torque assistance. Low interface stiffness and uncompensated dynamics can cause inefficient power delivery to the user, alter their perceived assistance and comfort, and destabilize feedback controllers. We hypothesize that the low stiffness in lateral strap interfaces, like those in the opensource M-BLUE exoskeleton, can be improved by mechanical redesign. Building on the open-source M-BLUE exoskeleton, this paper introduces an alternative interface design that loads the leg through anterior and posterior pads (normal loading) and straps, in which the pads provide extension assistance and the straps provide flexion assistance. We compare the interface dynamics of lateral and normal loading interfaces on N = 10 human subjects using both quasi-static spring measurements and frequency response methods, finding the new design to be 85.7% stiffer $(p<0.01)$ for a range of leg poses and in both flexion and extension loading.

  PublisherOA PDFDOI

• The clinical effects of the Össur Power Knee with phase-based and default control during sitting, standing, and walking

  Journal of NeuroEngineering and Rehabilitation · 2025-09-29 · 1 citations

  articleOpen accessSenior author

  BACKGROUND: A lack of evidence of compelling clinical benefits is a key factor limiting the adoption of commercialized powered robotic knee prostheses into mainstream clinical practice. Previous studies have demonstrated mixed results, potentially due to a combination of limitations in prosthetic hardware, control algorithms, and testing methodologies. METHODS: ) with n=7 above-knee amputee participants. Participants with both higher (K4) and lower mobility (K3) completed a series of experiments including repeated sitting and standing, a stand, walk, sit shuttle test, and fast walking on a treadmill. We tested both standard (ÖSSR) and novel (HKIC) control policies and compared the resulting clinical metrics to those found with the users' prescribed passive prostheses. Our experiments were physically demanding, which could help elucidate the potential benefits of powered knees. RESULTS: The clinical effects of the Power Knee varied with mobility level and the control policy used. The phase-based controller often produced stronger walking and sit/stand improvements for the higher mobility group compared to the default controller, though it also presented a steeper learning curve and reduced walk-to-sit transition speed. Conversely, the default control policy was perceived as easier to master but was less assistive to the higher mobility group and produced slower sit/stand cycles. Lower mobility participants experienced improvements in standing speed (HKIC: [Formula: see text]% faster, [Formula: see text]; ÖSSR: [Formula: see text]% faster, [Formula: see text]), inter-limb ground reaction force symmetry (HKIC: [Formula: see text], [Formula: see text]; ÖSSR: [Formula: see text], [Formula: see text]), and inter-limb peak knee moment symmetry (HKIC: [Formula: see text], [Formula: see text]; ÖSSR: [Formula: see text], [Formula: see text]) during sit-to-stand tasks relative to their passive prostheses. In contrast, higher mobility participants benefited less in sit/stand but showed improvements while walking including increased toe clearance (HKIC: [Formula: see text] mm, [Formula: see text]; ÖSSR: [Formula: see text] mm, [Formula: see text]), greater early stance knee flexion (HKIC: [Formula: see text], [Formula: see text]; ÖSSR: [Formula: see text], [Formula: see text]), and, for the HKIC policy, a reduced swing-phase peak hip flexion moment (HKIC: [Formula: see text] Nm/kg/(m/s), [Formula: see text]). Despite these biomechanical improvements and qualitative reports of reduced effort, neither control policy produced significant benefits in endurance or repeated task performance compared to the passive condition. Sit-to-stand cycle count in the lower mobility group was unchanged (HKIC: [Formula: see text], ÖSSR: [Formula: see text]), and it was reduced in the higher mobility group with the ÖSSR condition ([Formula: see text] fewer, [Formula: see text]). In the shuttle walk test, laps completed by higher mobility users decreased with HKIC ([Formula: see text] fewer, [Formula: see text]), and no significant differences were found for lower mobility users. No significant changes in fast walking distance or speed were observed across conditions. CONCLUSIONS: The latest generation Power Knee can create clinical improvements in walking and sit/stand behaviors compared to passive (microprocessor) knees, though the effects are sensitive to the user's mobility level and the Power Knee's control policy. However, these improvements did not directly translate to improved functional performance or endurance. Some negative effects of the Power Knee were also observed including reduced agility, slower transitions, and thermal limitations, though some of these limitations could potentially be addressed through future control innovations or with more thorough acclimation. The observed benefits motivate future longitudinal studies to investigate the clinical effects of robotic knees compared to passive (microprocessor) knees in real-world settings and to elucidate how they could be best utilized in clinical practice. TRIAL REGISTRATION: The experimental protocol was approved by the University of Michigan Institutional Review Board (HUM00230065) on February 9th, 2024. The trial is registered with the National Institutes of Health under ClinicalTrials.gov ID NCT06138977.

  PublisherOA PDFDOI

• A Task-Agnostic Approach to Unified Multi-Activity Gait Phase Estimation via Bilateral Sensing

  2025-05-12 · 2 citations

  articleOpen accessSenior author

  Estimating the gait phase is a key aspect for controlling many lower-limb rehabilitation robots, including transfemoral prostheses. Current control approaches often rely on high-level activity classification to then employ a taskspecific phase algorithm, which can limit adaptability across tasks and introduce risks associated with misclassification. This study proposes a novel unified phase variable framework with two approaches, one using activity classification and one being entirely task-agnostic. The framework uses predicted gait event information to continuously define a unified phase variable across level walking, ramp ascent/descent, and stair ascent/descent at various inclines and speeds. The classification approach senses the unilateral thigh angle, whereas the taskagnostic approach expands sensing to include the contralateral thigh angle. Simulated evaluations using an able-bodied dataset demonstrate average phase root-mean-square error of 6.8% with classification and 6.3% in the task-agnostic mode. The bilateral task-agnostic approach notably performed the same or better than the unilateral classification-based approach, showing improved consistency across subjects and tasks, particularly during stair ascent. These results highlight the feasibility of task-agnostic gait phase estimation for prosthesis control, demonstrating performance comparable to task-specific models while removing reliance on activity classification.

  PublisherOA PDFDOI

• A Clinical Tuning Framework for Continuous Kinematic and Impedance Control of a Powered Knee-Ankle Prosthesis

  IEEE Journal of Translational Engineering in Health and Medicine · 2025-01-01 · 1 citations

  articleOpen accessSenior author

  Objective: Configuring a prosthetic leg is an integral part of the fitting process, but the personalization of a multi-modal powered knee-ankle prosthesis is often too complex to realize in a clinical environment. This paper develops both the technical means to individualize a hybrid kinematic-impedance controller for variable-incline walking and sit-stand transitions, and an intuitive Clinical Tuning Interface (CTI) that allows prosthetists to directly modify the controller behavior. Methods and procedures: Utilizing an established method for predicting kinematic gait individuality alongside a new parallel approach for kinetic individuality, we personalize continuous-phase/task models of joint impedance (during stance) and kinematics (during swing) using tuned characteristics exclusively from level-ground walking. To take advantage of this method, we developed a CTI that translates common clinical tuning parameters into model adjustments for the walking and sit-stand controllers. We then conducted a case study where a prosthetist iteratively tuned the powered prosthesis to an above-knee amputee participant in a simulated clinical session involving sit-stand transitions and level walking, from which incline/decline walking features were automatically calibrated. Results: The prosthetist fully tuned the multi-activity prosthesis controller in under 20 min. Each iteration of tuning (i.e., observation, parameter adjustment, and model reprocessing) took 2 min on average for walking and 1 min on average for sit-stand. The tuned behavior changes were appropriately manifested in the commanded prosthesis torques, both at the manually tuned tasks and automatically tuned tasks (inclines). Conclusion: The CTI leveraged able-bodied trends to efficiently personalize a wide array of walking tasks and sit-stand transitions, demonstrating the efficiency necessary for powered knee-ankle prostheses to become clinically viable. Clinical impact: This paper introduces a clinical tuning interface that simplifies the tuning process for multimodal robotic prosthetic legs, reducing the time required from several hours to just 20 minutes thus improving clinical feasibility.

  PublisherDOI

Recent grants

• CAREER: Recovering and Enhancing Natural Locomotion in Changing Conditions with Powered Lower-Limb Prostheses and Orthoses

  NSF · $396k · 2019–2023

• NRI: FND: COLLAB: An Open-Source Robotic Leg Platform that Lowers the Barrier for Advanced Prosthetics Research

  NSF · $193k · 2017–2019

• Controlling Locomotion over Continuously Varying Activities for Agile Powered Prosthetic Legs

  NIH · $4.3M · 2018–2028

• NRI: INT: Collaborative Research: An Open-Source Framework for Continuous Torque Control of Intuitive Robotic Prosthetic Legs

  NSF · $940k · 2020–2025

• PHASE-BASED CONTROL OF LOCOMOTION FOR HIGH-PERFORMANCE PROSTHESES AND ORTHOSES

  NIH · $2.3M · 2013–2018

Frequent coauthors

• Nikhil Divekar

  University of Michigan–Ann Arbor

  45 shared

• Gray C. Thomas

  University of Michigan–Ann Arbor

  45 shared

• Jianping Lin

  University of Michigan–Ann Arbor

  31 shared

• Vamsi Peddinti

  University of Michigan–Ann Arbor

  31 shared

• T. Kevin Best

  University of Michigan–Ann Arbor

  29 shared

• Cara Welker

  University of Colorado Boulder

  23 shared

• Edgar Bolívar

  21 shared

• Elliott J. Rouse

  University of Michigan–Ann Arbor

  20 shared

Education

• Ph.D., Electrical and Computer Engineering

  University of Illinois at Urbana-Champaign

  2010

• B.S., Electrical Engineering and Computer Sciences

  University of California Berkeley

  2006