Distributed Collision-Free Motion Coordination on a Sphere: A Conic Control Barrier Function Approach

Tatsuya Ibuki, Sean Wilson, Aaron D. Ames, Magnus EgerstedtIEEE Control. Syst. Lett 2020 · DOI

TL;DR

This paper introduces a novel method for coordinating the movement of multiple rigid bodies on a spherical surface while ensuring they avoid collisions. It achieves this by extending the concept of Control Barrier Functions (CBFs) to the unique geometry of a sphere, enabling distributed, safe motion planning.

What Problem Does This Paper Address?

Coordinating the movement of multiple robots or agents is crucial for many applications, from drone swarms to planetary exploration. While much research has focused on motion coordination in flat 2D or 3D Euclidean spaces, many real-world scenarios involve movement on curved surfaces, such as spheres. Examples include coordinating satellites in orbit, managing underwater vehicles, or controlling robotic manipulators. A key challenge in these scenarios is ensuring that the agents not only reach their desired formations or tasks but also avoid colliding with each other, especially when their movement is constrained to a spherical workspace. Existing collision avoidance methods often rely on potential functions that can lead to overly cautious behavior or are designed for simpler geometric spaces, making them unsuitable for spherical constraints.

Methods

The paper proposes a distributed collision-free motion coordination strategy for rigid bodies on a sphere. The core of their approach lies in adapting Control Barrier Functions (CBFs) to this spherical geometry.

Here's a breakdown of their methodology:

  • Spherical Motion Modeling: They first model the motion of rigid bodies on a sphere. Crucially, they show that motion coordination on a sphere is equivalent to attitude coordination on the 3D Special Orthogonal group, SO(3). This mathematical insight allows them to leverage existing tools for attitude control.
  • Conic Control Barrier Functions (CBFs) on SO(3): To handle the spherical constraints, they develop a new type of CBF, termed "conic CBFs." These CBFs are designed to enforce angle-based constraints on the attitude of the rigid bodies, which directly translate to constraints on their positions on the sphere. They specifically address conic constraints, which define regions on the sphere.
  • Extension to Relative Motion and Collision Avoidance: The conic CBFs are then extended to handle relative motion between rigid bodies. This allows them to define a "safe set" for the entire network of rigid bodies, where safety is defined by maintaining a minimum geodesic distance between any two bodies.
  • Distributed Optimization: Each rigid body in the network computes its control input by solving a distributed optimization problem. This problem aims to achieve a predefined motion coordination task (e.g., attitude synchronization) while simultaneously satisfying the collision avoidance constraints imposed by the conic CBFs. This distributed nature means each robot only needs information about its neighbors, not a global view.

Datasets

Not discussed in the paper. The evaluation is performed via simulation.

Key Insights and Ideas

  • Sphere = SO(3) for Motion Coordination: The key insight is that motion coordination on a sphere can be elegantly mapped to attitude coordination on SO(3). This transformation simplifies the problem by allowing the use of powerful tools developed for rotational dynamics.
  • Conic CBFs for Spherical Safety: Developing "conic CBFs" is a significant contribution. These functions effectively translate geometric constraints on a sphere (like maintaining a minimum distance) into the language of CBFs, which are well-suited for guaranteeing safety in dynamical systems.
  • Distributed Control for Scalability: The use of a distributed optimization framework for control input generation is crucial. It means the system can scale to a larger number of agents without requiring a central controller, making it more practical for real-world deployments.
  • Balancing Task Achievement and Safety: The optimization formulation elegantly balances achieving a desired motion coordination task (e.g., flocking) with the imperative of maintaining safety (collision avoidance).

Key Results

The paper demonstrates the effectiveness of their proposed method through simulations.

  • Collision Avoidance Guarantee: The simulations show that the minimum geodesic distance between any pair of rigid bodies never drops below the defined collision distance ($D_c$), confirming that collision avoidance is achieved.
  • Cohesion and Alignment: The rigid body network successfully exhibits cohesion and alignment behaviors, indicating that the nominal motion coordination strategy is being pursued effectively while safety is maintained.
  • Distributed Implementation: The control inputs are computed in a distributed manner, relying only on information from neighboring agents, which is essential for practical implementation.

Limitations and Future Work

  • Robustness to Disturbances: The authors acknowledge that future work should include robustness analysis of their proposed conic CBF approaches against disturbances, as tackled in related work [31, 32].
  • Assumption on Initial Conditions: Theorem 2, which guarantees collision avoidance, assumes that collisions do not occur at the initial time. While this is a standard assumption for safety guarantees, real-world systems might start in unsafe configurations.
  • Geodesic Distance Assumption: The assumption $D_c < (\rho\pi)/2$ is made to avoid certain edge cases related to antipodal points on the sphere. While reasonable, it implies limitations on the size of the agents relative to the sphere's circumference.
  • Nominal Control Strategy: The performance of the collision avoidance system is dependent on the chosen nominal control strategy ($W_{nom,i}$). The paper uses an attitude synchronization law as an example, but other strategies could be explored.

Summary updated

BibTeX
@article{ibuki2020distributed,
  author = {Tatsuya Ibuki and Sean Wilson and Aaron D. Ames and Magnus Egerstedt},
  title = {{Distributed Collision-Free Motion Coordination on a Sphere: A Conic Control Barrier Function Approach}},
  journal = {IEEE Control. Syst. Lett},
  year = {2020},
  doi = {10.1109/lcsys.2020.2997952},
}
Figure