Planning for Tabletop Object Rearrangement

TL;DR

This work proposes an enhanced A*-based algorithm that improves state representation and employs incremental goal attempts with lazy evaluation at each iteration to enhance scalability while maintaining solution quality.

Abstract

Finding an high-quality solution for the tabletop object rearrangement planning is a challenging problem. Compared to determining a goal arrangement, rearrangement planning is challenging due to the dependencies between objects and the buffer capacity available to hold objects. Although orla* has proposed an A* based searching strategy with lazy evaluation for the high-quality solution, it is not scalable, with the success rate decreasing as the number of objects increases. To overcome this limitation, we propose an enhanced A*-based algorithm that improves state representation and employs incremental goal attempts with lazy evaluation at each iteration. This approach aims to enhance scalability while maintaining solution quality. Our evaluation demonstrates that our algorithm can provide superior solutions compared to orla*, in a shorter time, for both stationary and mobile robots.

Figures (9)

• Figure 1: An example of mobile robot tabletop rearrangement, while black arrows are the desired rearrangement goals.
• Figure 2: Left: A rearrangement task involves a table as the boundary box. The colored circles represent objects in their initial positions, while the dotted circles show their target positions. Right: Its dependency graph.
• Figure 3: Different definitions of travel cost bewtween $p_1$ and $p_2$. For stationary robots, the $TC$ is the red line length. For a mobile robot, given $b_1$ and $b_2$ which are the closest points to $p_1$ and $p_2$ on the table side, the $TC$ is the traveling distance moving along the table side (blue).
• Figure 4: Exploration process. The orange circles represent the states that are ready to be explored as in priority queue, whereas the blue circles indicate the states that have already been explored, and the green circle with $g$ is the state whose arrangement is the goal arrangement. In the upper section, during the exploration phase, the node with the minimum $f$ value is selected from the priority queue. The process of goal attempting involves searching for a feasible rearrangement solution from an explored state to the goal arrangement. In the lower section, if a plan exists that connects the start state to an explored state with a feasible solution to the goal, shown as purple path, and this plan is currently the best solution, its cost will be calculated. Subsequently, nodes in the priority queue with $f$ values exceeding this cost will be filtered out.
• Figure 5: The comparisons of different algorithms for stationary & mobile robots on different object number. The horizontal axis represents object number.