Fall 2026 GRASP SFI - Patrick Slade, Harvard University

In this talk

1. 4:09

   Personalizing robotic assistance

   Slade describes human-in-the-loop optimization: measure a person’s response to a robotic controller, then iteratively tune assistance to improve a chosen outcome. In a portable ankle exoskeleton study, wearable sensors and a data-driven model helped select assistance during natural walking; personalized settings roughly doubled the benefit of generic assistance for young, healthy participants.

2. 16:42

   Extending support across people and activities

   For people post-stroke, Slade’s collaboration with Connie Walsh’s group optimized ankle assistance to increase walking speed. Participants improved by about 0.1 meters per second, alongside gains in gait measures including symmetry and propulsion in the affected limb. In separate work, musculoskeletal simulation helped initialize assistance for varied tasks, while optimization adapted it to each person.

3. 29:19

   Adapting to movement goals

   Slade describes how people balance speed, stability, foot placement, and effort as they move. In a study varying walking speed, visual perturbation, and foot-placement targets across 27 conditions, participants’ movements helped reveal which goals they prioritized. Stability generally came first, followed by time and foot placement, with metabolic cost lower in the hierarchy.

4. 38:14

   Devices for demanding mobility needs

   Slade discusses a high-torque hip exoskeleton tested during steep-incline walking with 50-pound packs, and coordinated hip and ankle assistance that outperformed tuning each device alone. A linear actuator prototype reduced the second peak of knee contact force in participants with knee osteoarthritis. For blind and visually impaired users, the smartphone-based Mobilio navigation system performed similarly to a human guide and better than Google Maps in a user study.

5. 55:35

   Training, comfort, and choosing outcomes

   Slade says uncomfortable or unstable assistance can be excluded from optimization, while comfort may also be treated as an explicit goal. People continue adapting to devices over hours, and training design affects how well they learn to use assistance. For optimization, he recommends choosing high-level outcomes tied to user and clinician priorities, such as walking speed, rather than relying only on lower-level movement measures.