A Human-Machine-Environment Interactive Measurement System for Nonanthropomorphic Exoskeletons

Nonanthropomorphic structures have been increasingly used in lower limb exoskeletons. However, little research focuses on the human-machine-environment interactive measurement for nonanthropomorphic exoskeletons. This article presents a series of sensing methods on a nonanthropomorphic exoskeleton, which only uses six angular encoders at joints and an inertial measurement unit (IMU) on the back plate to observe the entire body motion state. The sensing system is designed to identify the walking terrain through kinematic calculation while mapping the angle of exoskeleton rods to that of human lower limbs. A five-segment dynamical model is built to estimate the ground reaction force (GRF). The interactive measurement system was designed, validated, and tested in experiments. We conducted walking experiments on different terrains. Gaits were collected on five road conditions: level ground, uphill and downhill, ascending stairs, and descending stairs. The differences between gaits on different terrain are analyzed and compared. The experimental results show that the terrain recognition method achieves the mean errors (MEs) of 0.003 +/- 0.021 and 0.0005 +/- 0.012 m for stride length and stride height, respectively. The gait angle mapping from exoskeleton rods to the wears' lower limbs achieves an overall root-mean-square error (RMSE) of 0.0236 +/- 0.0074 rad for thigh angle and 0.0277 +/- 0.0061 rad for shank angle. The GRF estimation achieves an overall RMSE of 0.1462 +/- 0.0255 body weight (BW) in the vertical direction and 0.0715 +/- 0.0187 BW in the anterior-posterior (AP) direction. The information on terrain, lower limb states, and GRFs can provide real-time feedback for the control of exoskeletons with similar nonanthropomorphic structures.