Motion Planning for Variable Topology Trusses: Reconfiguration and Locomotion

Chao Liu, Sencheng Yu, Mark Yim

Abstract

Truss robots are highly redundant parallel robotic systems that can be applied in a variety of scenarios. The variable topology truss (VTT) is a class of modular truss robots. As self-reconfigurable modular robots, a VTT is composed of many edge modules that can be rearranged into various structures depending on the task. These robots change their shape by not only controlling joint positions as with fixed morphology robots, but also reconfiguring the connectivity between truss members in order to change their topology. The motion planning problem for VTT robots is difficult due to their varying morphology, high dimensionality, the high likelihood for self-collision, and complex motion constraints. In this paper, a new motion planning framework to dramatically alter the structure of a VTT is presented. It can also be used to solve locomotion tasks that are much more efficient compared with previous work. Several test scenarios are used to show its effectiveness. Supplementary materials are available at https://www.modlabupenn.org/vtt-motion-planning/.

Introduction

Variable Topology Trusses (VTTs) are a versatile class of self-reconfigurable robots composed of linear actuator "edge modules" and connection "nodes." Unlike traditional fixed-skeleton robots, a VTT can alter both its geometry—by extending its members—and its fundamental topology by detaching and reattaching members through Split and Merge actions.

A physical prototype of a Variable Topology Truss (VTT) robot demonstrating its modular structure.

Planning motion for these systems is exceptionally challenging. They are highly redundant, often possessing over 20 degrees of freedom, and are prone to self-collision. Furthermore, they must adhere to strict physical constraints: the structure must remain rigid, nodes must maintain sufficient connectivity for control, and the center of mass must stay within the support base to prevent tipping. Standard sampling-based planners often fail because valid paths are hidden within narrow passages of a massive search space.

Methodology

To overcome these complexities, we propose a hierarchical framework that partitions the robot's configuration space and utilizes graph search for topological transitions:

  1. Dimensionality Reduction via Grouping: Nodes are grouped into pairs, and the "free space" for these groups is pre-computed by treating other members as static obstacles, drastically reducing the search space.
  2. Configuration Space Partitioning: The free space of a node is often divided into several disconnected "enclosed subspaces" by the surrounding truss members.
  3. Topology Planning as Graph Search: Moving a node between enclosed subspaces necessitates a change in topology. We use a Dijkstra-based graph search where the nodes represent these enclosed subspaces and the edges represent valid Split and Merge actions.
Visualization of the configuration space partitioned into enclosed subspaces (orange regions) used for planning topological transitions.
  1. Hardware-Informed Constraints: The planner accounts for realistic physical limits, including actuator lengths and joint angle limits.
  2. Non-Impact Locomotion: For rolling tasks, the planner expands the support polygon by moving internal nodes before tipping, ensuring the center of mass is always supported and avoiding damaging impacts.

Results and Evaluation

The framework was evaluated on specific test scenarios, including "Cube to Tower" reconfiguration and "Octahedron Rolling" locomotion. In benchmark tests against standard algorithms from the Open Motion Planning Library (OMPL), such as RRTConnect and TRRT, our proposed planner achieved a 100% success rate in scenarios where standard planners often dropped to 0-40% or took significantly longer.

In locomotion tasks, the sampling-based approach demonstrated remarkable efficiency. For an octahedron rolling task, it took an average of 1.18 seconds, compared to 166 seconds for previous optimization-based methods—a nearly 140x speedup. The framework also successfully planned complex multi-step topology changes that previously required manual waypoint intervention.

Discussion and Future Work

While the current framework provides a robust solution for static stability, future work will explore incorporating dynamic stability (inertia and momentum) for faster locomotion. Additionally, the current centralized approach may face scaling challenges; moving toward decentralized, "swarm-like" planning logic could enable the control of VTTs with hundreds or thousands of modules. Automating the optimal grouping of nodes for specific tasks also remains an open research challenge.

Summary updated

Abstract

Truss robots are highly redundant parallel robotic systems that can be applied in a variety of scenarios. The variable topology truss (VTT) is a class of modular truss robots. As self-reconfigurable modular robots, a VTT is composed of many edge modules that can be rearranged into various structures depending on the task. These robots change their shape by not only controlling joint positions as with fixed morphology robots, but also reconfiguring the connectivity between truss members in order to change their topology. The motion planning problem for VTT robots is difficult due to their varying morphology, high dimensionality, the high likelihood for self-collision, and complex motion constraints. In this paper, a new motion planning framework to dramatically alter the structure of a VTT is presented. It can also be used to solve locomotion tasks that are much more efficient compared with previous work. Several test scenarios are used to show its effectiveness. Supplementary materials are available at https://www.modlabupenn.org/vtt-motion-planning/.
BibTeX
@article{liu2023motion,
  author = {Chao Liu and Sencheng Yu and Mark Yim},
  title = {{Motion Planning for Variable Topology Trusses: Reconfiguration and Locomotion}},
  journal = {IEEE Trans. Robotics},
  year = {2023},
  doi = {10.1109/tro.2022.3228400},
}
Figure