Real-time trajectory planning based on joint-decoupled optimization in human-robot interaction. (February 2020)

• Record Type: Journal Article
• Authors: Zhang, Shiyu; Zanchettin, Andrea Maria; Villa, Renzo; Dai, Shuling
• Is Part Of: Mechanism and machine theory. Volume 144(2020)
• Publication Date: 2020-02
• Subjects: Real-time trajectory planning -- Human-robot interaction -- Non-linear optimization -- Machine learning -- Serial robot; Machine theory -- Periodicals; Machinery -- Periodicals; Machines -- Périodiques; Génie mécanique -- Périodiques; Machine theory; Machinery; Periodicals; 621.81
• DOI: 10.1016/j.mechmachtheory.2019.103664
• Languages: English
• ISSNs: 0094-114X

Abstract

Highlights: A real-time optimization-based trajectory planning method for robots is presented. The calculation time for optimization is reduced by more than one order magnitude. An acceptable solution can be got if the problem does not have a feasible solution. The solver can be scaled to problems with higher-dimensional DOFs. The method is implemented and validated in a specific human-robot interaction case. Abstract: In order to perform safe and natural interactions with humans, robots are required to adjust their motions quickly according to human behaviors. Performing the complex calculation and updating the trajectories in real-time is a particular challenge. In this paper, we present a real-time optimization-based trajectory planning method for serial robots. We encode the trajectory planning problem into a series of optimization problems. To solve the high-dimensional complex non-linear optimization problems in real-time, we provide a joint-decoupling method that transforms the original joint-coupled optimization problem into multiple joint-independent optimization problems, with much lower computational complexity. We implement and validate our method in a specific human-robot interaction case. Experimental results show that the computational feasibility and efficiency of optimization solution were greatly improved by the joint-decoupling transformation. Smooth, safe, and rapid motion of the robot was generated in real-time, establishing a basis for safe and reactive human-robot interactions.