Trajectory Planning of Slider-Pushers in Cluttered Environments with Automatic Switching

Non-prehensile manipulation presents a paradigm for manipulating objects that aims to extend the versatility of robotic tasks beyond conventional grasping. Integrating motion primitives like pushing, tipping, rolling, throwing, and sliding into robots brings forth a set of challenges related to sensing, control, and planning. In particular, planning becomes notably intricate when dealing with densely arranged obstacles. This work focuses on the act of pushing an object through a series of obstacles within a cluttered environment. In such a setting, it may become necessary to change the direction of the pushing action in order to navigate through narrow passages. Incorporating these directional switches introduces a symbolic level of decision making casting the problem in the task and motion planning framework. In this work we introduce a novel optimization-based algorithm tailored to a pusher-slider system, capable of handling highly constrained regions. This algorithm automatically determines a switching sequence and the corresponding trajectories. The infinite motion optimization problem is transcribed to a nonlinear program using B-splines. Our algorithm solves the problem jointly optimizing switching and trajectory parameters to establish a feasible path through the environment. Directional switches are introduced based on the properties of the intermediate solution. The work is demonstrated through simulation experiments in various cluttered environments.