Unobtrusive Gait Reconstructions Using Radar-Based Optimal Control Simulations

Radar-based movement analysis has the potential to overcome current limitations of optical or inertial motion capture since it is unobtrusive and contact-free. So far, it has only been used to estimate step length and cadence from radar. In this work, we aim to explore if we can also reconstruct gait kinetics and kinematics through radar-based optimal control. To create optimal control simulations, we solved a musculoskeletal dynamics trajectory optimization problem to find a motion that minimizes the difference to measured data while being physically plausible. In our radar-based optimal control approach, we simulate range-Doppler maps using a radar ray-tracing simulator and minimize the difference between the simulated and target range-Doppler maps. To validate our approach, we created synthetic range-Doppler maps of walking at 1.3 m/s and investigated if we could reconstruct the original walking gait cycle. Our gait cycle reconstruction showed a walking speed error of 0.08 m/s. For the hip and knee angles, we achieved a mean absolute error of 5.6° and 5.5°, which is in the same range as achieved by IMU-based optimal control. The ankle angle error was higher at 16.1°, which could have been caused by suboptimal placement of the radar antenna. Therefore, we conclude that reconstruction accuracy of radar-based optimal control is promising, yet needs to be improved.