In this project, autonomous locomotion planning and impedance control strategies will be designed and implemented on Exo-H3 with two purposes: 1) online shaping of personalized walking trajectory to enhance human comfort and 2) real-time adjustment of the exoskeleton impedance (flexibility) during human-robot interaction (HRI). Accordingly, intelligent control strategies will be developed for lower-limb exoskeletons (LLEs) to facilitate a balance between exoskeleton autonomy and human safety, which is technically challenging due to the non-passive and unpredictable behaviors of humans that make the detection of their intention and ensuring the flexibility of exoskeleton’s response harder.

The control strategies in this research project will advance the trending interdisciplinary research on various lower-limb pHRI tasks such as movement therapies, assistive locomotion, and human behavior assessment augmented by an exoskeleton. Accordingly, this research will advance the field of assistive robotics to have a widespread impact on the quality of life for a variety of people with disabilities and neurological conditions caused by SCI, Stroke, and other injuries/diseases. 

Current students focused on developing and implementing dynamic movement primitives (DMP) based motion generation with impedance control techniques to enhance adaptability, stability, and gait efficiency.  Another aim of this project is to create an intelligent controller utilizing adaptive central pattern generation (ACPG) control and reinforcement learning (RL) to improve postural stability and create personalized gait patterns.

In another intelligent control strategy, we modified the hip and ankle trajectories in real-time by minimizing divergent component of motion (DCM) error using time-independent DCM- based control strategies (TIC) and time-dependent DCM-based control (TDC) strategies to improve the postural stability of LLEs. These novel strategies yielded an average of a 16% improvement in postural stability, defined by the metrics observed, and a 35% improvement in real-time DCM error between the baseline testing and the four DCM-based strategies. 

Team members: Troy Schmidt, Zachary Knesek, Ashwin Unnikrishnan, Brandon Cisneros, Harshini Jayaprakash, Edward Lin, Sabrina Hsu, Eric Kwan