We are developing novel designs and model-based control algorithms for lower-limb and upper-limb robotic exoskeletons and prostheses using integrated biomechatronic systems modelling. Environment recognition systems and onboard intelligent sensors inform the dynamic controllers, which optimize the assistive loads and electrical energy regeneration capabilities of these wearable biomechatronic devices for geriatric and rehabiliation patients.
We are developing machine learning algorithms to automatically track, model, and control multibody dynamic systems. Our algorithms are deployed on mobile devices and control hardware units. Applications include markerless tracking for human movement biomechanics, control of autonomous vehicles, automotive powertrain models, and environment recognition systems for robotic lower-limb exoskeletons and prostheses.
We are developing model-predictive controller (MPC) algorithms to integrate planning and motion control of autonomous vehicles. These algorithms are deployed using robot operating system (ROS) on the University of Waterloo “Autonomoose”. Dynamic vehicle and tire models are developed from track tests using our intelligent vehicle measurement systems, and the controllers are evaluated in virtual simulations and hardware-in-loop experiments prior to vehicle testing.
We are developing physics-based models of various automotive components and subsystems, including tires, suspensions, engines, catalytic converters, batteries, drivebelts, and torque converters. Model reduction methods are used to create “control-oriented” models for model-predictive controllers. For hybrid electric and fuel cell powertrains, we are using machine learning algorithms to develop neural network models from experimental CAN measurements during vehicle testing on our track or $10M Green and Intelligent Automotive (GAIA) Research Facility.
Alongside human neuromusculoskeletal models, we have developed dynamic models and optimal controllers for upper-limb stroke rehabilitation robotic systems, which provide assistive/resistive loads according to patient-specific needs. We are now gamifying the system to enhance patient experience and are using artificial intelligence algorithms to learn patient-specific characteristics. The rehabilitation robotic system is currently being tested at Grand River Hospital.
We have developed high-fidelity dynamic models of golfer biomechanics, flexible shafts, ball flight aerodynamics, and clubhead-ball contact dynamics. Optimization methods are used to evaluate the effects of swing and equipment changes, in collaboration with golf club companies and professional regulatory bodies. Model validation is provided by motion capture systems, our AboutGolf simulator, pressurized air cannon, and high-speed video camera (up to 600,000 fps).
We work alongside Team Canada athletes (both Olympic and Paralympic) to improve their performance and equipment design. Predictive biomechanical models and optimal controllers were recently used to optimize the standing start biomechanics for Team Canada track cyclists. For Canada’s Wheelchair Curling Team, we designed and prototyped a 3D-printed novel curling end-effector that provides better control while delivering the curling stone. Predictive biomechanical models and optimal controllers of Team Canada Wheelchair Basketball players were used to analytically determine optimal seat positions that maximized propulsive forces.