Development of a novel foot slip sensor algorithm
Open Access
- Graduate Program:
- Mechanical Engineering
- Degree:
- Doctor of Philosophy
- Document Type:
- Dissertation
- Date of Defense:
- December 09, 2014
- Committee Members:
- Henry Joseph Sommer Iii, Dissertation Advisor
- Henry Joseph Sommer Iii, Committee Chair
- Christopher Rahn, Committee Member
- Stephen Jacob Piazza, Committee Member
- Jacob Willem Langelaan, Committee Member
- Jason Zachary Moore, Committee Member
- Keywords:
- slip sensor
- gait cycle detection
- walking robot
- stick-slip
- stochastic state estimation
- anomaly detection
- Abstract:
- A novel gait and slip detection algorithm for walking robots and humans using an inertial measurement unit (IMU) was developed. An unscented Kalman filter was formulated with a simple dynamic model of a foot as a block on a slope without translations. Considerable prediction errors resulted when unmodeled dynamics (i.e., translation) occurred. These prediction errors were systematically incorporated in a binary Bayes filter to estimate the probability of gait and slip states. Three stages of experiments with increasing complexities served dual purposes to develop and validate unified gait and slip detection algorithms. Multiple floor conditions and slopes were used in these experiments. Stick-slip experiments served as fundamental slip detection test cases without involving gait states. Monopedal walker experiments were used to develop the gait and slip detection algorithms for robots walking in “robot-style” static gait. Human subject experiments served as gait and slip detection test cases with typical “human-like” dynamic gait. Successful detection of gait and slip in all validation experiments was achieved using the same algorithms with commercial-grade IMU sensors. Continuous gait cycles were detected in proper order. Stance phase was successfully detected regardless of foot slip. Slip detection was successful except for very mild slips involving small jerk. One set of thresholds was used for detection under all floor conditions in a given series of validation experiments. It was not necessary to provide explicit knowledge of the inertial and control parameters, walking surface (orientation, contamination, and friction mode), or sensor error models. Consequently, a universal, simple-to-use, and robust realtime sensor algorithm was developed to detect gait and slip of robots walking on floors with various contaminations and slope.