Dr. Sugar and I are proud to be able to share Episode 19 of the Exoskeletons and Wearable Robotics Podcast for ExR’s Patreon Supporters! This time, we are lucky to have as a guest Dr. Jason Gillette, who shares his impressions of occupational exoskeletons. His Biomedical Engineering degree, combined with decades of hands-on expertise in ergonomics and multiple industrial exoskeleton projects, has given him a deep understanding of the field.

During this conversation, he is happy to share his observations on evaluating and task-matching industrial exoskeletons and wearable devices. Together, we discuss tools and methodologies for recording and analyzing job sites, from simple phone recording to EMG sensor studies and everything in between. The conversation flows between biomechanics and simple observations, underscoring the importance of understanding both objective and subjective data when conducting pilots.

All episodes are made available early for paid Patreon subscribers. Episode 19 is now published for everyone on YouTube and all major streaming platforms:

Main Topics: 00:00 Introduction 13:24 What are some tasks that make a lot of sense for occupational/industrial exoskeleton? 17:40 Can we have a digital tool that helps with exoskeleton to task matching? 22:37 TLV Equations (TLV = Threshold Limit Value, In this context it’s the Upper Limb Localized Fatigue TLV used to relate %MVC (percent of maximum voluntary contraction)) and duty cycle for raised-arm workand ways to calculate if a job is injury inducing? 30:38 Lessons learned from trials at companies like Toyota and John Deere 36:10 Additional hints or tips on exoskeleton to task matching. 40:08 Current work on evaluationing occupational exoskeletons in a medical setting for patient handling. 49:03 Observations on adoption of physical assistive technology. 55:05 Call to action: be more open to sharing field and lab data.

Sample of Dr. Gillette’s Work:

• Electromyography-based fatigue assessment of an upper body exoskeleton during automotive assembly – PubMed
• Artificial Intelligence for Injury Prevention (Professional Safety, Dec 2024): link

Links: Wearable Robotics Association: https://wearablerobotics.com

If you enjoy this content, join the physical revolution on Patreon / exoskeletonreport. Special thanks to our Patreon supporters for helping make this episode a reality! Editing: Borislav “Bobby” Marinov, Volunteer Producer: George Woodland, Title Card by Inna Marinova.

AI Generated Summary of Episode 19:

The episode dives into how to choose the right occupational exoskeleton for the right task, grounding the discussion in years of field and lab studies. The central theme: not every job benefits (from an exo or other ergonomic solution), and success hinges on matching assistance to posture, load, and exposure (how often the motion occurs).

The discussion starts with early, practical evaluations using welding and painting simulators. Performance with and without shoulder exoskeletons was comparable, but workers could sustain performance longer with assistance, an early hint that the devices “take the edge off” rather than transform capability.

From there, the conversation moves to what’s happening “under the skin.” Wireless EMG became the backbone method for comparing muscle activation with and without exoskeletons, first in controlled lab simulations and then in demanding field contexts at major manufacturers. The key was apples-to-apples comparisons on the same tasks, repeated with and without the device.

A recurring question is whether a statistically significant change is also meaningful. The approach adopted uses localized muscle fatigue Threshold Limit Values (TLVs) that combine intensity (% of capacity) and duty cycle (time exposed). Mapping tasks to these curves turns EMG data into “fatiguing vs. non-fatiguing” decisions.

Field results informed clear “sweet spots” for shoulder support. Typical passive/assistive designs that engage roughly from 60° to 135° of arm elevation performed best when work clustered around 90°–135°, involved intermediate hand loads (about 3–7 lb tools/parts), and occurred frequently (e.g., half the shift). Benefits often shrink with very light loads unless the movement is highly repetitive; conversely, very heavy tools exceed what shoulder exos can reasonably offset.

Because large factories map tasks down to seconds, predictive models became possible. By linking posture bands, tool weights, and frequencies to observed EMG reductions and TLV classifications, hundreds of jobs can be triaged on paper to identify the best exoskeleton candidates before investing in on-floor trials. That triage approach accelerates pilots and focuses attention on the highest-yield tasks.

Usability and perception matter as much as EMG. Satisfaction in the field often splits into thirds: enthusiastic, neutral, and opposed. Two factors correlate strongly with acceptance: perceived impact on job performance (it must not slow anyone down) and body conformance (the device should hug the body and “disappear” when not actively assisting). EMG and perceived exertion usually align in the sweet spot, but diverge at the edges—e.g., extreme overhead may feel brutal even when EMG changes look modest—so both quantitative and subjective measures are necessary.

To scale task assessment beyond specialists, the episode explores using a phone camera and AI pose tracking to extract joint angles from work videos, combine that with entered tool/part weights, and automatically classify fatigue risk and likely exoskeleton benefit. Validation against lab motion capture and EMG was promising, but field complexity (crowds, lifts, occlusions) challenged robustness. The open question is whether this lives as a standalone app or integrates into broader ergonomic suites.

Standardization is another pillar. A developing test method (under a standards body) uses a configurable fixture to evaluate shoulder exoskeletons across representative postures (including flexion and abduction) and tool loads (e.g., 0/3/5/7 lb). The output ties EMG deltas to TLVs to delineate where assistance reliably shifts tasks from fatiguing to non-fatiguing, alongside parallel tracking of perceived exertion and usability.

The discussion then widens to backs, necks, and static or awkward postures outside automotive assembly. Examples include construction logistics with multi-level shelf work, prolonged leaning in clinical or dental settings, and highly repetitive sorting. A staged research program in patient handling focuses on wheelchair-to-bed and bed-to-wheelchair transfers using different back exoskeleton designs, instrumented with EMG, motion checks, and usability ratings. Early lab findings show promise; next steps involve short, carefully controlled clinical pilots that account for infection control and device hygiene.

Task matching remains the everyday bottleneck, and the advice is pragmatic. Start with classic ergonomic analyses, listen to workers about where they feel shoulder vs. back strain, and bring in a second opinion (consultants/integrators) to avoid deploying the wrong device and souring first impressions. Fit and setup time matter; few people will swap devices mid-shift, so blended demands (e.g., mixed shoulder and back loading) may call for hybrid strategies or different process design.

The episode closes with a call for more transparent, longitudinal evidence. Day-to-day, task-level data, alongside honest reporting of where exoskeletons don’t help, would speed iteration, inform smarter devices, and, most importantly, keep the focus on protecting workers so they can “age gracefully” (and retire injury-free) in their professions.

Many thanks to YOU, our Patreon supporters, for encouraging us and offsetting some of the cost of making and editing this podcast! Additional thanks to Dr. Gillette for taking hours of his time and for his willingness to share his experience and know-how. Dr. Sugar for being my trusty cohost, our Volunteer Producer, George Woodland, and Inna Marinova for making the title card graphic.