A Distributed Pipeline for Scalable, Deconflicted Formation Flying (2003.01851v2)

Abstract: Reliance on external localization infrastructure and centralized coordination are main limiting factors for formation flying of vehicles in large numbers and in unprepared environments. While solutions using onboard localization address the dependency on external infrastructure, the associated coordination strategies typically lack collision avoidance and scalability. To address these shortcomings, we present a unified pipeline with onboard localization and a distributed, collision-free motion planning strategy that scales to a large number of vehicles. Since distributed collision avoidance strategies are known to result in gridlock, we also present a decentralized task assignment solution to deconflict vehicles. We experimentally validate our pipeline in simulation and hardware. The results show that our approach for solving the optimization problem associated with motion planning gives solutions within seconds in cases where general purpose solvers fail due to high complexity. In addition, our lightweight assignment strategy leads to successful and quicker formation convergence in 96-100% of all trials, whereas indefinite gridlocks occur without it for 33-50% of trials. By enabling large-scale, deconflicted coordination, this pipeline should help pave the way for anytime, anywhere deployment of aerial swarms.

• The paper presents a distributed pipeline for UAV swarms that integrates onboard VIO and decentralized control to achieve scalable, collision-free formation flying.
• The approach improves formation convergence rates to 96-100% in both simulations and hardware tests compared to 33-50% with traditional methods.
• The research highlights the benefits of decentralization by eliminating the need for external positioning systems and enabling robust, real-world UAV deployments.

Overview of "A Distributed Pipeline for Scalable, Deconflicted Formation Flying"

This paper introduces a distributed pipeline aimed at enabling scalable and collision-free formation flying for swarms of unmanned aerial vehicles (UAVs). The research targets two critical challenges in the operational deployment of large-scale vehicle swarms: autonomous localization without reliance on external infrastructure and coordination strategies that simultaneously ensure collision avoidance and scalability. The authors propose an integrated approach combining onboard localization with distributed formation control and a task assignment strategy to address these issues effectively.

Primary Contributions

The paper's contributions are notable across several areas of multi-robot systems and swarm robotics:

1. Scalable Formation Control: The authors extend existing formation control frameworks by integrating them with onboard localization systems. This approach effectively negates the need for external positioning systems, utilizing visual inertial odometry (VIO) for localization. The control strategy is based on scaled rotation matrices that promote robustness against pose errors, disturbances, and misalignments.
2. Distributed Task Assignment: Employing an auction-based method, the authors present a distributed task assignment strategy that effectively deconflicts vehicle paths, solving gridlocks often caused by distributed collision avoidance strategies. The task assignment does not require a common reference frame and operates on local information, demonstrating excellent scalability.
3. Demonstrations and Simulations: The pipeline's effectiveness is validated through both simulation and hardware experiments, showcasing the potential for real-world applications. The authors demonstrate substantial improvements in formation convergence rates, with a particular focus on large-scale simulations involving up to 100 UAVs and successful operation without reliance on centralized coordination. Results indicate significant success, as convergence to the desired formation configuration occurs in 96-100% of trials, compared to 33-50% without the distributed assignment.

Theoretical and Practical Implications

• Optimization and Scalability: The research advances computational techniques for gain matrix optimization in formation control, presenting a scalable algorithm using the alternating direction method of multipliers (ADMM) that addresses the limitations of existing solvers for problems involving large numbers of vehicles.
• Robustness through Decentralization: By dismissing the need for common axes alignment and employing a robust distributed control method, this work highlights the practical robustness achievable through decentralization, thus reducing failure points associated with reliance on centralized systems.
• Practical Deployment: By using only onboard sensors and low-bandwidth V2V communication, this pipeline is poised to support real-world deployments of aerial swarms in varied environments, most notably where conventional GPS-based localization is infeasible.

Future Directions

The paper opens various avenues for further research. Potential extensions include incorporating vehicle dynamics in the assignment strategy to minimize travel distances more effectively and potentially integrating more advanced pose graph optimization techniques to enhance VIO pose estimates' consistency. Furthermore, exploring variations on distributed auction algorithms and optimizing their adaptation to dynamic environments could significantly enhance their in-field applications.

Conclusion

This research signifies advancement toward practical and large-scale deployment of UAV swarms capable of operating independently of external infrastructure. The proposed distributed pipeline enhances the formation flying capabilities of multirotor systems, highlighting promising implications for applications in both civil and defense domains. Herein lies an incremental yet critical advancement toward more autonomous and resilient multi-agent robotic systems.