My research is centered around acheiving dynamically stable cruthc-less exoskeleton walking on a fully-actuated lower-body exoskeleton, Atalante. Below are the various research projects that I’ve conducted with this robotic platform. You can also find a complete listing of my research on my google scholar profile .
Preference-Based Learning for Personalized Exoskeleton Gait Optimization
In this work, we developed a novel algorithm, CoSpar, that utilized methods from preference-elicitation to learn the underlying preference function of exoskeleton users. This framework was first verified in simulation, and then experimentally conducted for 3 able-bodied subjects. The experiments explored obtaining both 1 and 2 dimensional bayesian posteriors. The first publication of this work received Best Overall Paper at ICRA 2020, as well as Best Paper in Human-Robot Interaction. Later, we also extended this work to optimize over higher-dimensional action spaces using dimensionality reduction techniques.
- Tucker, M., Novoseller, E., Kann, C., Sui, Y., Yue, Y., Burdick, J., & Ames, A. D. (2019). Preference-Based Learning for Exoskeleton Gait Optimization. In 2020 IEEE International Conference on Robotics and Automation (ICRA), 2020. IEEE ICRA Best Overall Paper Award. IEEE ICRA Best Paper in Human-Robot Interaction Award.
- Tucker, M., Cheng, M., Novoseller, E., Cheng, R., Yue, Y., Burdick, J. W., & Ames, A. D. (2020). Human Preference-Based Learning for High-dimensional Optimization of Exoskeleton Walking Gaits. Under Review.
Towards Variable Assistance via Controlled Set Invariance
In this work, we proposed and demonstrated a method of acheiving variable assitance on the exoskeleton. This framework used tools from controlled set invariance, specifically control barrier functions. This framework was validated through two separate experiments. The results of these experiments showed that lower levels of exoskeleton assistence resulted in higher metabolic expenditure rates for 8 able-bodied subjects.
- Gurriet, T., Tucker, M., Duburcq, A., Boeris, G., & Ames, A. D. (2019). Towards Variable Assistance for Lower Body Exoskeletons. IEEE Robotics and Automation Letters, 5(1), 266-273.