New advanced wearable exosuits are capable of restoring function to individuals in the older adult population by reducing the metabolic cost of walking and restoring normative biomechanics. An important function of these devices is to timely and accurately recognize user intent and optimize the control to provide biomechanically appropriate assistance across multimodal task paradigms. Key challenges in the wearable robotics control community include generalizing control systems across a rich variety of real-world tasks and diverse individuals while simultaneously personalizing control systems to each individual’s specific set of biomechanical needs. Standard state-based controllers have largely failed to provide this level of task generalization combined with individualized assistance. A unified framework that focuses on continuous estimation of internal human state variables provides promise for task agnostic control that tracks individualized movements. This talk will examine our approaches for human internal state estimation using deep learning, optimization of the control, and associated human outcomes of the approach across multi-modal activities. New open-source datasets that we have generated to facilitate research in this area will also be briefly discussed. 

In this talk, I will present our recent solution for personalized robotic prosthetic leg control that not only achieves the desired prosthesis mechanics but also augments unamputated biological joint motion to facilitate human-robot coordination during walking. At the core of our solution is a new bilevel optimization framework, which incorporates iterative inverse reinforcement learning (IRL) and reinforcement learning (RL) approaches to personalize the human-prosthesis objective function and prosthesis control law. Our preliminary validation demonstrated the feasibility of this bilevel optimization framework for symbiotic prosthesis control design.

Wearable robots are commonly used to provide basic functionality such as mediating standing and level-ground walking. To achieve robust control of these devices during widely-varying ambulation tasks, hierarchical control systems have been implemented that classify an individual’s high-level intention using discrete labels and delegate the phase-varying joint-level response of the device within a predicted mode to mid-level controllers that render motion, torque or impedance reference trajectories. This presentation will highlight our efforts to eliminate this hierarchy by developing volitional and semi-volitional control systems of robotic knee-ankle prostheses using new sensing modalities of peripheral muscles and shared robot control paradigms. Secondly, this presentation will emphasize that each device wearer is unique in their physical form as well as the priority they place on specific neuromotor objectives during movement. Predictive neuromusculoskeletal modeling systems that incorporate an individual’s anthropometric shape and underlying task objectives will be discussed within the context of soft-hip flexion exosuits.

We have developed vision-based systems to assist with navigational challenges in unstructured environments. One system uses RGB-D video data to detect staircases, estimate the distance to the last step on level ground, and estimate the height and depth of each tread. The difference in perspective when descending versus ascending stairs results in lower accuracy for downstairs estimates. Another system uses RGB images to detect anomalies in the path ahead. An autoencoder is trained on obstacle-free sidewalk images, so that anomalous objects produce image reconstruction errors. Subsequent classifiers discriminate common “anomalies” like manhole covers from novel obstacles. These systems can be used by assistive robots to adapt system behavior to match environmental conditions, and by vision-impaired users to warn of upcoming conditions.