Soft Exosuits vs. Rigid Frames: A New Era for Rehabilitation Engineering

1. Introduction: Redefining the Human-Machine Interface

The history of wearable robotics has long been dominated by the visual and mechanical language of the exoskeleton: rigid, anthropomorphic frames of metal and carbon fiber, powered by heavy electric motors or hydraulics, designed to envelop the human limb and force it into motion. This design philosophy, popularized by science fiction and pursued vigorously by engineering labs for half a century, operates on the principle that to assist the human skeleton, one must essentially replicate it. While this approach has yielded significant successes in enabling complete paraplegics to stand and walk, it has hit a persistent "metabolic wall" when applied to individuals with residual mobility—such as soldiers, stroke survivors, or the elderly. For these populations, the added mass, joint misalignment, and kinematic restrictions of rigid systems often increase the metabolic cost of walking rather than reducing it. The robot becomes a burden to be carried rather than a partner to be utilized.

In response to these fundamental limitations, a radical divergence in robotic design emerged from the Harvard Biodesign Lab at the John A. Paulson School of Engineering and Applied Sciences and the Wyss Institute for Biologically Inspired Engineering. Led by Professor Conor Walsh, this research group introduced the concept of the "Soft Robotic Exosuit." The premise was deceptively simple: if the user has a functioning skeletal structure, the robot does not need to provide a rigid parallel skeleton. Instead, it needs to provide tensile forces that mimic the function of muscles and tendons. By removing the rigid frame and transmitting force through functional textiles, the exosuit eliminates the inertia and restriction of traditional robots, creating a system that is transparent to the user when unpowered and synergistic when active.

This report provides an exhaustive analysis of the development, engineering principles, control strategies, and translational applications of soft robotic exosuits at Harvard University. It traces the technology's evolution from early DARPA-funded prototypes aimed at military load carriage to FDA-cleared medical devices for stroke rehabilitation and commercially deployed industrial safety systems. The analysis explores the complex interplay between textile engineering, biomechanics, and human-in-the-loop optimization that allows these systems to break the metabolic cost paradox and redefine the future of human augmentation.

1.1 The Metabolic Cost Paradox of Rigid Systems

To understand the necessity of the soft exosuit, one must first understand the limitations of its rigid predecessors. The human body is an incredibly efficient walking machine, utilizing pendulum mechanics and elastic energy storage in tendons to minimize fuel consumption. When a rigid exoskeleton is strapped to the legs, several parasitic effects emerge. First, the "added mass penalty" is severe; adding mass to the distal segments of the legs (shanks and feet) increases the moment of inertia, requiring the hip muscles to work significantly harder to swing the leg.¹ Second, rigid joints rarely align perfectly with human biological joints. The human knee, for example, is not a simple pin joint; it rolls and slides. A rigid robot hinge that does not match this complex trajectory creates internal shear forces, fighting the user's natural motion and causing discomfort or injury.² Finally, rigid systems impose kinematic constraints, often locking degrees of freedom that are not being assisted, which forces the wearer into an unnatural, "robotic" gait that is metabolically expensive.

The Harvard soft exosuit approach circumvents these issues by utilizing the wearer's own skeleton to bear compressive loads. The robot applies only tensile forces, parallel to the muscles. This dramatically reduces the weight of the wearable portion of the device—often to less than a kilogram for the textile components—and places the heavier actuation components (motors and batteries) near the body's center of mass (the waist or lower back).³ This architecture minimizes the metabolic penalty of the device's weight, allowing the assistive power to yield a net metabolic benefit.

1.2 The Harvard Biodesign Philosophy

The guiding philosophy of the Harvard Biodesign Lab is "synergy." Unlike autonomous robots that operate in isolation, wearable robots must function as a seamless extension of the human body. This requires a deep integration of disparate fields: apparel design, biomechanics, control theory, and mechatronics. Professor Conor Walsh and his team recognized early on that a mechanical engineer alone could not build a comfortable exosuit; it required the expertise of textile scientists to understand how fabric deforms under load and biomechanists to identify the precise timing of muscle recruitment.³

This interdisciplinary approach led to the "Un-Exoskeleton" concept. The device is designed to be worn like clothing, utilizing breathable, flexible materials that do not restrict movement. When the device is slack, it exerts zero impedance on the wearer, feeling essentially like a pair of pants. This "transparency" is a critical safety and usability feature, as it allows the user to move freely in confined spaces or perform agility maneuvers without fighting the machine.⁶ The transition from the "Iron Man" suit to the "Power-Dressed" athlete represents a fundamental shift in how we conceptualize human-machine interaction, moving from dominance and replacement to cooperation and augmentation.

2. Engineering Architecture: The Soft-Hard Interface

The engineering challenge of the soft exosuit is non-trivial: how does one transmit hundreds of Newtons of force through a soft, compliant interface (human skin and fat) without causing pain, slippage, or energy loss? The solution developed by the Harvard team involves a sophisticated architecture of functional textiles, cable-driven actuation, and soft sensing.

2.1 Functional Textiles and Load Paths

The most visible innovation of the soft exosuit is the textile interface. Standard fabrics stretch and deform unpredictably under high loads. To transmit assistive force effectively, the textile must be engineered with specific "load paths"—lines of high-stiffness material (such as webbing or specialized weaves) that route the force from the actuator to the skeletal anchor points.⁶

2.1.1 Anatomy of the Textile Anchor

The exosuit typically relies on two primary anchor points: the waist and the calf/foot.

• The Waist Belt: This component serves as the proximal anchor. It must handle the reaction forces generated when the suit pulls on the leg. If the belt is not secure, the force intended to extend the hip will instead pull the belt down, causing discomfort and loss of efficiency. The Harvard design utilizes a multi-layered structure involving rip-stop nylon, foam padding (polyurethane or EVA), and webbing straps to distribute the shear forces over a large surface area of the pelvis.⁷

• The Calf/Boot Interface: This distal anchor transmits the force to the foot for ankle plantarflexion assistance. The design often integrates with the user's footwear or uses a specialized calf wrap. The challenge here is to prevent the cuff from migrating up the leg when pulled. The lab developed "gait-synchronized" tightening mechanisms and high-friction inner surfaces to ensure the textile locks onto the limb during the active phase of the gait cycle.⁷

2.1.2 The Human-Textile Series Elasticity

A critical biomechanical feature of the textile design is its inherent compliance. In robotics, "Series Elastic Actuators" (SEAs) are often used to introduce a spring between the motor and the load to absorb shocks and improve force control. In the soft exosuit, the textile itself acts as the series spring. When the motor pulls the cable, the textile stretches slightly before the force ramps up on the human limb. This compliance is safer for the biological tissue, preventing sudden, jarring spikes in force that could cause injury. However, it introduces a control challenge: the system must model and predict this stretch to deliver the correct force at the correct time.⁶

2.2 Actuation Mechanics: The Bowden Cable Paradigm

While early soft robotics research explored pneumatic artificial muscles (McKibben actuators), the Harvard team largely standardized on electromechanical Bowden cable drives for their lower-limb exosuits. Pneumatics, while inherently soft, suffer from low energetic efficiency (due to compressibility of air and thermodynamic losses) and require bulky compressors that are difficult to miniaturize for mobile applications.³

2.2.1 The Bowden Cable Advantage

The Bowden cable architecture allows for the physical separation of the power source (motor) and the point of application (joint).

• Proximal Mass Concentration: The motors and batteries—the heaviest components—are mounted on the waist or in a backpack.

• Distal Lightness: Only the lightweight cables and textile attachments are placed on the legs.This configuration is crucial for reducing the metabolic cost of walking. Biomechanics research shows that adding 1kg to the feet increases metabolic cost far more than adding 1kg to the waist. By keeping the distal segments light, the exosuit avoids the "pendulum penalty" that plagues rigid exoskeletons.9

2.2.2 Actuator Specifications

The actuation units typically employ high-torque, brushless DC motors connected to a spool mechanism. The motor retracts the inner cable relative to the outer sheath, creating tension.

• Force Capabilities: These actuators are capable of generating forces exceeding 200 Newtons, which translates to significant assistive torques at the ankle and hip joints (e.g., 30% of biological moments).¹⁰

• Bandwidth: The system must be fast enough to match the rapid changes in muscle activity during walking. The Harvard team’s custom actuators improved bandwidth from ~6Hz in early prototypes to over 20Hz in later iterations, ensuring the device can respond instantly to gait events.¹³

• Mobile vs. Off-Board: The research has utilized two distinct platforms. "Off-board" systems use powerful motors mounted on a cart or frame next to a treadmill, transmitting power via long cables. This setup removes weight constraints, allowing researchers to probe the physiological limits of assistance. "Portable" systems carry everything on the body, introducing the constraint of battery life and weight trade-offs, which is essential for real-world validation.¹⁴

2.3 Sensor Fusion and the "Soft" Observer

A rigid robot can use high-precision optical encoders at its joints to know exactly where the limb is. A soft exosuit, having no rigid joints, poses a unique sensing challenge: how to track the human body's position without encumbering it?

2.3.1 Inertial Measurement Units (IMUs)

The primary sensing modality for the Harvard exosuits involves IMUs placed on the thigh and shank segments. These sensors contain accelerometers and gyroscopes that measure the angular velocity and acceleration of the limb segments.

• Kinematic Reconstruction: Algorithms fuse the data from these IMUs to estimate the joint angles (hip and knee flexion/extension) in real-time. This allows the controller to know the phase of the gait cycle (e.g., "mid-stance" vs. "terminal swing").⁹

2.3.2 Soft Sensors and Force Feedback

To enable closed-loop force control, the suit monitors the tension in the cables using load cells. Furthermore, the lab has pioneered the development of hyper-elastic strain sensors—silicone-based sensors filled with conductive liquid metals or carbon particles—that can be embedded directly into the textile. These sensors stretch with the fabric, providing direct feedback on the suit's interaction with the body without introducing rigid hard points.¹⁷

2.3.3 Gait Event Detection

Precise timing is the cornerstone of soft exosuit assistance. The system must detect "Heel Strike" (initial contact) and "Toe Off" to synchronize its actuation. The Harvard team developed robust, threshold-based algorithms that analyze the gyroscope signals from the shank to detect these events with millisecond precision, even across varying walking speeds and terrains.¹⁶ This reliability is critical; if the suit triggers too early or too late, it can trip the user or increase metabolic cost.

3. Control Strategies: The Human-in-the-Loop Revolution

The hardware of a soft exosuit is useless without an intelligence to drive it. The control problem is fundamentally different from industrial robotics. In a factory, the robot dictates the motion. In an exosuit, the human dictates the motion, and the robot must assist transparently. The Harvard group's most significant contribution to the field may be their advancement of "Human-in-the-Loop" (HIL) optimization strategies, which acknowledge that the human and the machine are a coupled, co-adapting system.

3.1 The Challenge of Inter-Subject Variability

Early experiments revealed a frustrating truth: a control profile that significantly helped one participant might hinder another. Differences in muscle strength, tendon stiffness, limb length, and neural control strategies meant that "average" parameters were often suboptimal for the individual. For example, the optimal timing for the onset of ankle assistance might vary by 10-20% of the gait cycle between users.¹⁹ This variability necessitated a move away from fixed control laws toward adaptive, personalized algorithms.

3.2 Bio-Inspired Force Profiles

The baseline control strategy mimics the biological torque profiles of the muscles being assisted.

• Ankle Assistance: The suit is programmed to remain slack during the early stance phase (to allow for shock absorption and stability). As the leg moves into terminal stance, the suit ramps up tension, mimicking the eccentric and concentric contraction of the Soleus and Gastrocnemius muscles. This provides a "push-off" boost.²

• Hip Assistance: For hip extension, the suit detects the moment of maximum hip flexion (when the leg is furthest forward) and begins to pull the thigh backward as the foot hits the ground, simulating the action of the gluteus maximus.²¹

3.3 Optimization Algorithms: Closing the Loop with Physiology

To solve the personalization problem, the Harvard team implemented HIL optimization, where the user's physiological response (measured via metabolic cost or muscle activity) serves as the feedback signal for the controller.

3.3.1 Bayesian Optimization

The team adapted Bayesian Optimization for this task. This machine learning approach is particularly well-suited for biological signals, which are often "noisy" (high variability) and "expensive" to collect (requiring minutes of walking to get a stable metabolic reading).

• Exploration vs. Exploitation: The algorithm builds a probabilistic model (a Gaussian Process) of how the user's metabolic rate changes with different control parameters (e.g., peak force timing). It then intelligently selects the next set of parameters to test. It might choose a parameter set likely to be better (exploitation) or one where its knowledge is uncertain (exploration).

• Results: In comparative studies, Bayesian optimization was able to identify optimal control parameters for an individual in approximately 21 minutes of walking. This rapid convergence is crucial for clinical viability, as patients cannot walk for hours to tune a device.¹⁹

3.3.2 Covariance Matrix Adaptation Evolution Strategy (CMA-ES)

Another powerful algorithm employed is CMA-ES. This is a stochastic, derivative-free method that "evolves" a population of control parameters over generations. It is highly robust to the noise inherent in human physiological data. By iteratively updating a covariance matrix that defines the search distribution, CMA-ES can navigate the complex, non-convex landscape of human metabolic cost to find the global minimum.²⁴

3.4 Table 1: Comparison of Optimization Strategies

Feature	Gradient Descent	Bayesian Optimization	CMA-ES
Mechanism	Follows the slope of the cost function	Builds a probabilistic surrogate model	Evolves a population distribution
Data Efficiency	Low (requires many samples)	High (works with sparse data)	Medium (requires generations)
Noise Tolerance	Poor (can get stuck in local minima)	Good (handles uncertainty well)	Excellent (robust to outliers)
Convergence Time	Slow (> 45 mins)	Fast (~20 mins) 22	Medium
Use Case	Simple, low-noise systems	Clinical tuning, Metabolic optimization	Complex, multi-parameter spaces

4. Physiological Mechanisms and Metabolic Economies

The ultimate litmus test for any gait-assistive device is metabolic cost: does it reduce the amount of oxygen the wearer consumes? For decades, rigid exoskeletons struggled to break even. In 2017, the Harvard Biodesign Lab shattered this barrier with a landmark study published in Science Robotics, proving that soft exosuits could deliver massive metabolic savings.

4.1 The Science Robotics 2017 Breakthrough

In this study, seven load carriers walked on a treadmill carrying a load equivalent to 30% of their body weight. They wore a soft exosuit assisting the ankle joint.

• The Result: The study reported a 22.8% ± 3.17% reduction in metabolic cost relative to walking with the device powered off.²⁶

• Significance: This reduction is comparable to the effect of taking off a 15-20 pound backpack. It was the highest relative reduction reported for an autonomous wearable robot at that time.

• Mechanism: The study showed that the savings did not just come from the ankle. The assistance provided at the ankle had a positive downstream effect on the hip, reducing the effort required by the hip flexors to swing the leg. This demonstrated that local assistance could have global benefits for the entire kinetic chain.²⁶

4.2 Muscle-Tendon Interaction

Detailed electromyography (EMG) analysis reveals how the suit interacts with specific muscles.

• Soleus and Gastrocnemius: The suit acts in parallel with the calf muscles. When the suit pulls, the activation of the Soleus decreases significantly—by approximately 40% in some conditions.²⁰

• Metabolic Efficiency: Interestingly, the suit allows the biological muscle fibers to operate at more efficient lengths and velocities. By taking up the slack and providing the high-force "kick" at push-off, the suit spares the muscles from their most energetically expensive task.²⁸

4.3 Joint Synergy: Hip vs. Ankle

While the ankle contributes the most power to walking, the Harvard team also explored multi-joint assistance.

• Hip Extension: Assisting the hip alone provides a metabolic reduction of ~5-10%.

• Multi-Joint: Combining hip and ankle assistance yields greater savings, but not necessarily a linear sum. The study found that assisting all three joints (hip, knee, ankle) could reduce metabolic cost by up to 50% in emulator settings, but in portable soft suits, the sweet spot often lies in the hip-ankle combination, balancing complexity/weight with benefit.³⁰

4.4 Table 2: Metabolic and Biomechanical Outcomes

Metric	Outcome	Context/Study
Net Metabolic Reduction	22.8% ± 3.17%	Loaded walking, Ankle Assist 27
Soleus Muscle Activity	~40-50% Reduction	Peak activation during stance 20
Walking Economy (Hip-Only)	~5-8% Reduction	Hip extension assistance 21
Propulsion Asymmetry	~20% Improvement	Stroke patients (ReStore) 32

5. Medical Application: Clinical Translation to Stroke Rehabilitation

While the military applications focused on "superhuman" endurance, the medical track focused on restoration. Stroke is a leading cause of long-term disability, often leaving survivors with hemiparesis (one-sided weakness). The Harvard team realized that the soft exosuit's lightweight, non-restrictive nature made it an ideal tool for gait retraining.

5.1 Post-Stroke Gait Pathology

Hemiparetic gait is characterized by specific deficits:

• Drop Foot: Weakness in the dorsiflexors (tibialis anterior) causes the toes to drag during the swing phase, creating a tripping hazard.

• Propulsion Deficit: The paretic leg fails to push off effectively, forcing the healthy leg to compensate.

• Compensations: Patients often "hike" their hip or swing the leg in a circle (circumduction) to clear the ground, patterns that are inefficient and can lead to secondary orthopedic issues.³²

5.2 The ReStore System

The medical version of the technology, named ReStore, was developed in collaboration with ReWalk Robotics (now Lifeward). Unlike the military suit, the ReStore is designed for unilateral (one-legged) assistance.

• Dual-Action Cables: The ReStore features two cables. An anterior cable pulls the foot up (dorsiflexion) during the swing phase to prevent drop foot. A posterior cable pulls the heel up (plantarflexion) during the stance phase to assist with propulsion.³⁴

• Therapeutic Tool: The device is not intended as a permanent assistive device (like a prosthesis) but as a rehabilitative tool used in physical therapy clinics. It allows therapists to prolong training sessions and increase intensity without physically exhausting themselves manually moving the patient's limb.³⁵

5.3 Clinical Trial Results and FDA Clearance

Extensive clinical trials validated the efficacy of the ReStore system.

• Propulsion Symmetry: The suit significantly improved the propulsive symmetry between the paretic and non-paretic legs. By providing a boost to the weak leg, it encouraged the patient to trust and use that limb.³²

• Ground Clearance: The active dorsiflexion assistance reduced the need for hip hiking and circumduction, promoting a more natural, symmetrical gait pattern.³⁶

• Speed and Endurance: Patients using the system showed improvements in the 10-meter walk test (speed) and the 6-minute walk test (endurance).³⁸

In June 2019, the ReStore system received FDA 510(k) clearance, a historic milestone as the first soft robotic patient interface to be cleared for medical use in the United States.³⁹ This clearance validated the safety and efficacy of the soft exosuit approach, opening the door for its commercial adoption in clinics nationwide.

6. Industrial Application: Verve Motion and Logistics

Beyond the battlefield and the clinic, the Harvard team identified a third critical application: the warehouse. Back injuries are the number one cause of musculoskeletal disorders in the workplace, particularly in logistics and grocery distribution where workers lift thousands of pounds daily.

6.1 The "Industrial Athlete"

Warehouse workers are often referred to as "industrial athletes" due to the physical demands of their job. However, unlike sports athletes, they perform these feats for 8-10 hours a day, 5 days a week, often with poor biomechanics due to fatigue. Rigid industrial exoskeletons existed but faced adoption barriers: they were bulky, heavy, and could become snagging hazards in narrow aisles. They were also uncomfortable to wear while driving forklifts or sitting during breaks.

6.2 The Verve Motion Spin-out

In 2020, the technology was spun out into a startup, Verve Motion, to commercialize the SafeLift exosuit.

• Mechanism: The SafeLift is a soft, backpack-like system that runs cables parallel to the user's back muscles. Sensors detect when the worker enters a lifting posture (bending at the waist/hips). As the worker begins to lift, the motors retract the cables, offloading the lumbar spine.⁴¹

• Performance: The suit provides up to 240 Newtons of lift assistance, which effectively offloads 40% of the weight of the object being lifted from the user's back.¹² For a worker lifting 50,000 pounds cumulatively over a shift, this saves the back from bearing 20,000 pounds of load every single day.

6.3 Economic Impact and Safety Outcomes

Verve Motion's deployment has generated compelling real-world data.

• Injury Reduction: In pilot studies involving over 65 million lifts across various industries (grocery, retail, logistics), the use of the SafeLift exosuit resulted in a 60-85% reduction in back injury rates.⁴²

• Productivity: Contrary to fears that safety gear slows workers down, the fatigue reduction provided by the suit often leads to a productivity boost of 4-8%, as workers maintain their energy levels throughout the shift.⁴³

• Adoption: The "soft" form factor is the key differentiator. Workers can wear the suit all day, sit in vehicles, and navigate tight spaces without removing it, leading to high compliance rates compared to rigid alternatives.⁴⁴

7. Challenges, Limitations, and Future Directions

Despite the successes, the field of soft exosuits faces ongoing challenges that define the current research agenda.

7.1 The Comfort and Interface Barrier

While "softer" than metal, high-tension textiles can still cause discomfort.

• Shear Forces: Transmitting 200N through the skin can cause chafing or bruising if the fit is not perfect. The "migration" of the textile (sliding up or down the limb) remains a difficult engineering problem, requiring constant adjustment or custom fitting.⁴⁵

• Thermal Burden: To support these loads, the suits often involve significant wrapping of the limbs with foam and webbing. This impedes sweat evaporation, leading to thermal discomfort, especially in hot industrial or military environments. User feedback consistently highlights heat management as a barrier to long-term acceptance.⁴⁶

7.2 Actuation Bottlenecks

The Bowden cable system, while effective, is mechanically inefficient due to friction between the cable and the sheath. It also requires a rigid motor box.

• Future Actuators: The Harvard lab is actively investigating next-generation actuators that are fully soft. This includes fluidic systems and electro-hydraulic actuators that contract like biological muscle tissue. These systems could theoretically eliminate the rigid motor pack entirely, distributing the actuation throughout the textile.¹⁷

7.3 The AI Horizon (2025 and Beyond)

As of 2025, the focus of the Harvard Biodesign Lab has shifted toward deeper integration of Artificial Intelligence.

• Predictive Control: Current systems react to the gait cycle. Future systems aim to predict user intent before movement occurs, using machine learning to analyze muscle twitch (EMG) or subtle shifts in center of pressure.

• Aging-in-Place: A major new frontier is assisting the aging population. Lightweight, AI-driven soft apparel could provide the subtle assistance needed to help the elderly stand from a chair or walk with confidence, potentially delaying the need for assisted living.⁴⁸

8. Conclusion

The development of the Soft Robotic Exosuit at Harvard University represents a seminal moment in the history of robotics. It challenged the prevailing dogma that robotic assistance required rigid, skeletal structures. By proving that functional textiles and cable-driven actuation could effectively transmit force, reduce metabolic cost, and restore gait symmetry, Conor Walsh and his team opened a new branch of the phylogenetic tree of machines.

From the high-performance demands of the DARPA Warrior Web program to the restorative needs of stroke survivors and the daily grind of the industrial athlete, the soft exosuit has proven its versatility. Its success lies in its humility; it does not attempt to overpower or replace the human body, but to understand and cooperate with it. As the technology matures, integrating more advanced materials and AI-driven personalization, the boundary between "clothing" and "robot" will continue to dissolve, leading to a future where wearable assistance is as ubiquitous, comfortable, and essential as the shoes on our feet.

Table 3: Evolution of Harvard Soft Exosuit Technology (2012-2025)

Era	Focus	Key Technology	Primary Milestone
2012-2014	Proof of Concept	Pneumatics / Early Cables	DARPA Warrior Web Funding; "Un-Exoskeleton" idea 49
2015-2017	Optimization & Physiology	Multi-joint Bowden Cables	Science Robotics paper (23% metabolic reduction) 26
2018-2020	Clinical Translation	ReStore System	FDA Clearance for Stroke Rehab 40
2021-2024	Industrial Scale	SafeLift (Verve Motion)	Commercial scaling; Injury reduction data 42
2025+	AI & Ubiquity	Machine Learning / Soft Actuators	Integration with home care & aging population 48

(Note: Citations such as refer to the specific research snippets utilized to substantiate the facts presented in this report.)

Works cited

1. Soft exosuit economies: understanding the costs of lightening the load - Wyss Institute, accessed January 2, 2026, https://wyss.harvard.edu/news/soft-exosuit-economies-understanding-the-costs-of-lightening-the-load/

2. Harvard engineers design soft robotic exosuits to aid human movement, accessed January 2, 2026, https://www.harvardmagazine.com/2013/12/wearable-robots

3. Conor Walsh - Progressive Engineer, accessed January 2, 2026, https://progressiveengineer.com/conor-walsh/

4. The $3 million suit - Harvard Gazette, accessed January 2, 2026, https://news.harvard.edu/gazette/story/2014/09/the-3-million-suit/

5. The robot suit providing hope of a walking cure | Medical research | The Guardian, accessed January 2, 2026, https://www.theguardian.com/science/2016/nov/20/robot-exosuit-harvard-helps-stroke-sufferers-walk-again

6. Soft Exosuits | Harvard Biodesign Lab, accessed January 2, 2026, https://biodesign.seas.harvard.edu/soft-exosuits

7. US9351900B2 - Soft exosuit for assistance with human motion - Google Patents, accessed January 2, 2026, https://patents.google.com/patent/US9351900B2/en

8. EP2895133B1 - Soft exosuit for assistance with human motion - Google Patents, accessed January 2, 2026, https://patents.google.com/patent/EP2895133B1/fi

9. Systematic evaluation of human-exosuit physical interfaces - Vanderbilt University, accessed January 2, 2026, https://cdn.vanderbilt.edu/vu-my/wp-content/uploads/sites/1409/2016/03/14121849/DW_Final_Joint_Matthew_B_Yandell_v3.pdf

10. Biologically-inspired soft exosuit - PubMed, accessed January 2, 2026, https://pubmed.ncbi.nlm.nih.gov/24187272/

11. A biologically-inspired multi-joint soft exosuit that can reduce the energy cost of loaded walking, accessed January 2, 2026, https://dash.harvard.edu/bitstreams/7312037d-d1c7-6bd4-e053-0100007fdf3b/download

12. SafeLift - Exoskeleton Report, accessed January 2, 2026, https://exoskeletonreport.com/product/safelift/

13. Design of a Quasi-Direct Drive Actuator with Embedded Pulley for a Compact, Lightweight, and High-Bandwidth Exosuit - MDPI, accessed January 2, 2026, https://www.mdpi.com/2076-0825/12/1/21

14. A biologically-inspired multi-joint soft exosuit that can reduce the energy cost of loaded walking - NIH, accessed January 2, 2026, https://pmc.ncbi.nlm.nih.gov/articles/PMC4864923/

15. A soft exosuit for patients with stroke: Feasibility study with a mobile off-board actuation unit, accessed January 2, 2026, https://sites.bu.edu/nrl/files/2022/06/2015_-_bae_-_icorr_-_a_soft_exosuit_for_patients_with_stroke._feasibility_study_with_a_mobile_off-board_actuation_unit.pdf

16. A Novel Gait Event Detection Algorithm Using a Thigh-Worn Inertial Measurement Unit and Joint Angle Information - ASME Digital Collection, accessed January 2, 2026, https://asmedigitalcollection.asme.org/biomechanical/article/146/4/044502/1193869/A-Novel-Gait-Event-Detection-Algorithm-Using-a

17. Soft Robotics - Harvard Biodesign Lab, accessed January 2, 2026, https://biodesign.seas.harvard.edu/soft-robotics

18. Inertial sensors for gait monitoring and design of adaptive controllers for exoskeletons after stroke: a feasibility study - PMC - NIH, accessed January 2, 2026, https://pmc.ncbi.nlm.nih.gov/articles/PMC10513467/

19. Human-in-the-loop optimization of hip assistance with a soft exosuit during walking - PubMed, accessed January 2, 2026, https://pubmed.ncbi.nlm.nih.gov/33141683/

20. Enhanced gastrocnemius-mimicking lower limb powered exoskeleton robot - PMC, accessed January 2, 2026, https://pmc.ncbi.nlm.nih.gov/articles/PMC12323188/

21. IMU-based Iterative Control for Hip Extension Assistance with a Soft Exosuit, accessed January 2, 2026, https://biodesign.seas.harvard.edu/publications/imu-based-iterative-control-hip-extension-assistance-soft-exosuit

22. Human-in-the-loop Bayesian optimization of wearable device parameters | PLOS One, accessed January 2, 2026, https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0184054

23. Human-in-the-loop Bayesian optimization of wearable device parameters - Harvard DASH, accessed January 2, 2026, https://dash.harvard.edu/bitstreams/7312037e-82d5-6bd4-e053-0100007fdf3b/download

24. Covariance Matrix Adaptation Evolution Strategy - Algorithm Afternoon, accessed January 2, 2026, https://algorithmafternoon.com/strategies/covariance_matrix_adaptation_evolution_strategy/

25. CMA-ES - Wikipedia, accessed January 2, 2026, https://en.wikipedia.org/wiki/CMA-ES

26. Soft exosuit economies: understanding the costs of lightening the load - Harvard School of Engineering and Applied Sciences, accessed January 2, 2026, https://seas.harvard.edu/news/2017/01/soft-exosuit-economies-understanding-costs-lightening-load

27. Assistance magnitude versus metabolic cost reductions for a tethered multiarticular soft exosuit, accessed January 2, 2026, https://regmedia.co.uk/2017/01/25/scienceroboticsarticle.pdf

28. Soft Robotic Exosuit - YouTube, accessed January 2, 2026, https://www.youtube.com/watch?v=azSpdF8CGPw

29. Individualization of exosuit assistance based on measured muscle dynamics during versatile walking - PMC - NIH, accessed January 2, 2026, https://pmc.ncbi.nlm.nih.gov/articles/PMC9052350/

30. Comparing optimized exoskeleton assistance of the hip, knee, and ankle in single and multi-joint configurations - NIH, accessed January 2, 2026, https://pmc.ncbi.nlm.nih.gov/articles/PMC10936256/

31. Comparing optimized exoskeleton assistance of the hip, knee, and ankle in single and multi-joint configurations - bioRxiv, accessed January 2, 2026, https://www.biorxiv.org/content/10.1101/2021.02.19.431882v1.full.pdf

32. ReStore Soft Exo-Suit for gait rehabilitation | NICE, accessed January 2, 2026, https://www.nice.org.uk/guidance/mib239/resources/restore-soft-exosuit-for-gait-rehabilitation-pdf-2285965580878789

33. Scientific research with stroke patients - Harvard Biodesign Lab, accessed January 2, 2026, https://biodesign.seas.harvard.edu/scientific-research-stroke-patients-testing

34. ReStore™ Soft Exo-Suit For Stroke Rehabilitation - Lifeward, accessed January 2, 2026, https://golifeward.com/products/restore-exo-suit/

35. The ReWalk ReStore™ soft robotic exosuit: a multi-site clinical trial of the safety, reliability, and feasibility of exosuit-augmented post-stroke gait rehabilitation - PMC - NIH, accessed January 2, 2026, https://pmc.ncbi.nlm.nih.gov/articles/PMC7301475/

36. O 089 - A soft robotic exosuit assisting the paretic ankle in patients post-stroke: Effect on muscle activation during overground walking - PubMed, accessed January 2, 2026, https://pubmed.ncbi.nlm.nih.gov/29983296/

37. Effects of a soft robotic exosuit on the quality and speed of overground walking depends on walking ability after stroke - PubMed Central, accessed January 2, 2026, https://pmc.ncbi.nlm.nih.gov/articles/PMC10474762/

38. Investigation of the Effectiveness of the Robotic ReStore Soft Exoskeleton in the Development of Early Mobilization, Walking, and Coordination of Stroke Patients: A Randomized Clinical Trial - MDPI, accessed January 2, 2026, https://www.mdpi.com/2218-6581/13/3/44

39. 510(k) Premarket Notification - FDA, accessed January 2, 2026, https://www.accessdata.fda.gov/scripts/cdrh/cfdocs/cfpmn/pmn.cfm?ID=K190337

40. FDA Issues Clearance for the ReStore™ Exo-Suit, the First Soft Robotic System for Stroke Therapy | Lifeward Ltd., accessed January 2, 2026, https://ir.rewalk.com/news-releases/news-release-details/fda-issues-clearance-restoretm-exo-suit-first-soft-robotic

41. SafeLift | Powered Soft Exosuit For Workers DEMO - Verve Motion, accessed January 2, 2026, https://learn.vervemotion.com/safelift-powered-soft-exosuit-for-workers-demo

42. Real Stories of Verve Motion's SafeLift™ Exosuit in the Workplace, accessed January 2, 2026, https://vervemotion.com/customer-results/

43. Verve Motion releases study on soft exosuits - Robotics 24/7, accessed January 2, 2026, https://www.robotics247.com/article/verve_motion_releases_study_on_soft_exosuits

44. FAQ - Verve Motion, accessed January 2, 2026, https://vervemotion.com/faq/

45. Soft robots built for extreme environments - OAE Publishing Inc., accessed January 2, 2026, https://www.oaepublish.com/articles/ss.2023.51

46. Thermophysiological aspects of wearable robotics: Challenges and opportunities: Temperature - Taylor & Francis, accessed January 2, 2026, https://www.tandfonline.com/doi/abs/10.1080/23328940.2022.2113725

47. Thermophysiological aspects of wearable robotics: Challenges and opportunities - PMC, accessed January 2, 2026, https://pmc.ncbi.nlm.nih.gov/articles/PMC10405755/

48. 2025 a2 National Symposium Plenary Talk: Conor Walsh, PhD — ML/AI in Soft Wearable Robots - MassAITC, accessed January 2, 2026, https://massaitc.org/2025/07/02/2025-a2-national-symposium-plenary-talk-conor-walsh-phd-machine-learning-and-ai-methods-for-aging-and-dementia-care/

49. Walsh to advance wearable robot design - Harvard School of Engineering and Applied Sciences, accessed January 2, 2026, https://seas.harvard.edu/news/2014/09/walsh-advance-wearable-robot-design