A multi-mode real-time terrain parameter estimation method for wheeled motion control of mobile robots. (1st May 2018)

• Record Type: Journal Article
• Authors: Li, Yuankai; Ding, Liang; Zheng, Zhizhong; Yang, Qizhi; Zhao, Xingang; Liu, Guangjun
• Is Part Of: Mechanical systems and signal processing. Volume 104(2018)
• Page Start: 758
• Page End: 775
• Publication Date: 2018-05-01
• Subjects: Terrain parameters -- Real-time estimation -- Multi-mode -- recursive Gauss-Newton method -- adaptive robust extended Kalman filter; Structural dynamics -- Periodicals; Vibration -- Periodicals; Constructions -- Dynamique -- Périodiques; Vibration -- Périodiques; Structural dynamics; Vibration; Periodicals; 621
• DOI: 10.1016/j.ymssp.2017.11.038
• Languages: English
• ISSNs: 0888-3270

Abstract

Highlights: A real-time terrain parameter estimation method for control use is developed. The method is hierarchical with a type of two-layer structure. The inner layer provides a rough real-time result for control use by using AREKF. The outer layer online optimize the inner layer result based on RGN algorithm. The method has three fundamental estimation modes: EKF, REKF and RGN. Multiple modes make the method applicable for flat, rough and non-uniform terrains. The wheel motion control accuracy can be improved effectively. Abstract: For motion control of wheeled planetary rovers traversing on deformable terrain, real-time terrain parameter estimation is critical in modeling the wheel-terrain interaction and compensating the effect of wheel slipping. A multi-mode real-time estimation method is proposed in this paper to achieve accurate terrain parameter estimation. The proposed method is composed of an inner layer for real-time filtering and an outer layer for online update. In the inner layer, sinkage exponent and internal frictional angle, which have higher sensitivity than that of the other terrain parameters to wheel-terrain interaction forces, are estimated in real time by using an adaptive robust extended Kalman filter (AREKF), whereas the other parameters are fixed with nominal values. The inner layer result can help synthesize the current wheel-terrain contact forces with adequate precision, but has limited prediction capability for time-variable wheel slipping. To improve estimation accuracy of the result from the inner layer, an outer layer based on recursive Gauss-Newton (RGN) algorithm is introduced to refine the result of real-time filtering according to the innovation contained in the history data. With the two-layer structure, the proposed method can work in three fundamental estimation modes: EKF, REKF and RGN, making the method applicable for flat, rough and non-uniform terrains. Simulations have demonstrated the effectiveness of the proposed method under three terrain types, showing the advantages of introducing the two-layer structure.