ReloPush: Multi-object Rearrangement in Confined Spaces with a Nonholonomic Mobile Robot Pusher

TL;DR

ReloPush tackles multi-object rearrangement in confined spaces using a nonholonomic mobile robot pusher. It introduces the Push-Traversability graph (PT-graph) that encodes geometric, kinematic, and stability constraints as pushing poses and Dubins-based transitions, enabling efficient planning. The framework greedily decomposes the task into single-object pushes, using prerelocations and blocking-object removals to maintain feasibility, and relies on Hybrid A* for reachability and Dubins planning for pushing. Through simulation and real-world experiments with up to nine blocks in a $4\times 5.2\mathrm{m}^2$ workspace on a 1/10-scale MuSHR, ReloPush achieves orders-of-magnitude faster planning and more robust execution than baselines without the PT-graph. This work advances practical pushing-based rearrangement under tight workspace constraints and suggests a path toward handling more unstructured objects.

Abstract

We focus on push-based multi-object rearrangement planning using a nonholonomically constrained mobile robot. The simultaneous geometric, kinematic, and physics constraints make this problem especially challenging. Prior work on rearrangement planning often relaxes some of these constraints by assuming dexterous hardware, prehensile manipulation, or sparsely occupied workspaces. Our key insight is that by capturing these constraints into a unified representation, we could empower a constrained robot to tackle difficult problem instances by modifying the environment in its favor. To this end, we introduce a Push-Traversability graph, whose vertices represent poses that the robot can push objects from, and edges represent optimal, kinematically feasible, and stable transitions between them. Based on this graph, we develop ReloPush, a graph-based planning framework that takes as input a complex multi-object rearrangement task and breaks it down into a sequence of single-object pushing tasks. We evaluate ReloPush across a series of challenging scenarios, involving the rearrangement of densely cluttered workspaces with up to nine objects, using a 1/10-scale robot racecar. ReloPush exhibits orders of magnitude faster runtimes and significantly more robust execution in the real world, evidenced in lower execution times and fewer losses of object contact, compared to two baselines lacking our proposed graph structure.

Figures (9)

• Figure 1: In this work, we describe ReloPush, a planning framework for tackling multi-object rearrangement tasks with a nonholonomic mobile robot pusher.
• Figure 2: The ReloPush architecture. Given the initial pose of the pusher and a rearrangement task in the form of start/goal object poses, ReloPush plans an efficient sequence of rearrangement subtasks to be executed by the robot via pushing.
• Figure 3: PT-graph generation. (\ref{['fig:ptgraph1']}) First, every object is assigned $K$ pushing poses (e.g., a cubic object has 4 pushing poses). (\ref{['fig:ptgraph2']}) For any pair of pushing poses, we check if a collision-free path that respects the steering limit for quasistatic pushing can be drawn. (c) For each valid path, we construct a directed edge between its start/goal vertices.
• Figure 4: (\ref{['fig:tgraph1']}) Two objects (navy squares) need to be rearranged to goal poses (yellow squares). (\ref{['fig:tgraph2']}) The PT-graph: nodes are pushing poses and edges are Dubins paths connecting them. By searching the graph, we can determine if any blocking objects need to be removed. For instance, the initial pose of object 1 is found to be blocking the shortest rearrangement of object 2 (red path).
• Figure 5: Two Dubins curves with the same goal pose (top right) and maximum turning radius. When the start pose is too close to the goal ($d\leq d_{th}$), the resulting Dubins curve (green color) is a Short Path involving large turns violating the workspace boundary. Using Dubins path classification shkel2001classificationlim2023circling, we can determine a prerelocation of the object's starting pose to allow reaching the goal via a Long Path ($d> d_{th}$) which will involve smaller turning arcs (gray color).

Theorems & Definitions (6)

• Theorem IV.1
• proof
• Theorem IV.2
• proof
• Theorem IV.3
• proof