Push Planning for Object Placement in Clutter Using the PR-2

The goal of this project is to investigate the implementation of a planning algorithm for the problem of placing objects on a cluttered surface with a PR-2 mobile manipulator. The original push planning algorithm [1] was initially developed as a simulation. We modified the simulator for execution in real-world cluttered environments. This paper discusses the challenges of implementation and presents empirical results that determine how well the simulator models the real world as clutter is pushed and collides with other objects. Within recent years, there has been more attention directed towards picking up and placing objects on surface with increased levels of natural clutter [1], [2], [3]. The presence of clutter is very prevalent in real-world environments. Algorithms that work well at grasping and placing objects in controlled environments, often fail once applied to a realworld scenario. The classical motion planners that work well in the laboratory either require sparsely populated tables in order to perform reliable picking and placing tasks or assume that other objects (clutter) are off limits and cannot be manipulated in order to accomplish a task. This does not scale well in the real-world, and does not resemble how humans interact with their environment on a daily basis. The push planning algorithm [1] implemented on the PR-2 in this paper addresses the problem of placing objects on a cluttered surface, where non prehensile push actions may be able to create space for the object. These actions may also result in collisions between objects that make up the clutter, which are modelled using a 2D physics engine. The algorithm has two main components. The first is a planner that selects a sequence of pushing actions to clear a space large enough to place the target object. The second is a heuristic that guides search by finding candidate placements that are likely to yield simpler manipulation plans. We investigated the implementation of this push planning algorithm on the Willow Garage PR-2 (Fig. 1a).