The Feasible Transition Graph: Encoding Topology and Manipulation Constraints for Multirobot Push-Planning

Abstract

Abstract. Planning for multirobot manipulation in dense clutter becomes particularly challenging as the motion of the manipulated object causes the connectivity of the robots' free space to change. This paper introduces a data structure, the Feasible Transition Graph (FTG), and algorithms that solve such complex motion planning problems. We define an equivalence relation over object configurations based on the robots' free space connectivity. Within an equivalence class, the homogeneous multirobot motion planning problem is straightforward, which allows us to decouple the problems of composing feasible object motions and planning paths for individual robots. The FTG captures transitions among the equivalence classes and encodes constraints that must be satisfied for the robots to manipulate the object. From this data structure, we readily derive a complete planner to coordinate such motion. Finally, we show how to construct the FTG in some sample environments and discuss future adaptations to general environments.

Constraint Manifold Subsearch for multirobot path planning with cooperative tasks

“…In the future, we will pursue techniques for automatically constructing the constraint manifold [6,9,24,27] to apply CMS to more types of tasks. Of special interest is the task of cooperatively carrying an object through an environment that is sufficiently cluttered that robots must disconnect and reconnect to the object clear obstacles [19].…”

Physics-based trajectory optimization for grasping in cluttered environments

“…Hauser et al [10] consider the problem of removing the minimum number of movable obstacles for the robot to achieve its objective. Lindzey et al [19] consider an extension where objects in the environment are pushed and rearranged by multiple robots. Kaelbling et al [13] present an integrated task and motion planner that plans backwards from the objective of grasping a desired target object using a regression-based symbolic planner.…”

More Shuttles, Less Cost: Energy Efficient Planning for Scalable High-Density Warehouse Environments

“…That is, it does not suffice to compute their trajectory. Instead, the planning robot must manipulate the environment itself (Ben-Shahar and Rivlin 1998; Stilman et al 2007;Lindzey et al 2014). Not all means of manipulating can be transferred to our application-e.g., grasping, sweeping and toppling as proposed by (Dogar and Srinivasa 2011)-or are unnecessarily complex (e.g., pushing).…”