Bimanual Robotic Shoe Lacing via Graphs of Convex Sets Motion Planning

Amr Marey, Mahdi Tavakoli, Qing Zhao

Keywords: Robotic manipulation, object manipulation, Motion planning, Optimization

TL;DR: This paper presents a bimanual robotic shoe-lacing system in the Drake simulation environment, leveraging workspace segmentation and Graphs of Convex Sets for precise motion planning with flexible objects.

Abstract: Robotic shoe lacing poses a significant challenge for autonomous manipulation due to the need for precise coordination, contact-rich interactions, and the handling of flexible objects within constrained geometries. This work presents a bimanual robotic system, modeled with two Franka Emika Panda arms in the Drake simulation environment, designed to address this benchmark task. The workspace, consisting of a shoe, shoelace, aglets, table, and eyelets, was developed using custom SDF assets to provide a realistic testbed. Iterative Regional Inflation by Semidefinite programming (IRIS) was applied to segment the environment into convex regions, which were then incorporated into a Graphs of Convex Sets (GCS) formulation for trajectory optimization. The resulting framework generated collision-free trajectories that successfully guided shoelace aglets through successive eyelets, achieving reliable execution in simulation. These results demonstrate the effectiveness of convex optimization–based methods for complex manipulation tasks and highlight the potential of structured planning approaches in advancing dexterous robotic autonomy.

Submission Number: 62