Online Trajectory Generation with Distributed Model Predictive Control for Multi-Robot Motion Planning

Abstract: We present a distributed model predictive control (DMPC) algorithm to generate trajectories in real-time for multiple robots. We adopted the \textit{on-demand collision avoidance} method presented in previous work to efficiently compute non-colliding trajectories in transition tasks. An event-triggered replanning strategy is proposed to account for disturbances. Our simulation results show that the proposed collision avoidance method can reduce, on average, around 50% of the travel time required to complete a multi-agent point-to-point transition when compared to the well-studied Buffered Voronoi Cells (BVC) approach. Additionally, it shows a higher success rate in transition tasks with a high density of agents, with more than 90% success rate with 30 palm-sized quadrotor agents in a 18 m^3 arena. The approach was experimentally validated with a swarm of up to 20 drones flying in close proximity.

• The paper introduces a Distributed Model Predictive Control (DMPC) framework for online multi-robot trajectory generation with on-demand collision avoidance.
• Testing shows the DMPC method reduces travel time by 50% and achieves over 90% success rates in dense multi-robot scenarios.
• This research enhances multi-robot capabilities for real-time tasks in dynamic environments, crucial for applications like logistics, surveillance, and autonomous navigation.

Online Trajectory Generation with Distributed Model Predictive Control for Multi-Robot Motion Planning

The paper "Online Trajectory Generation with Distributed Model Predictive Control for Multi-Robot Motion Planning" presents a novel framework for real-time trajectory generation in multi-robot systems, utilizing a Distributed Model Predictive Control (DMPC) strategy. The approach aims to address challenges related to dynamic environments where multiple decision-making agents operate within shared spaces to accomplish complex tasks. The research is carefully positioned within the context of existing optimization-based techniques, offering a substantial contribution towards scalable, efficient, and robust multi-robot trajectory generation.

Overview

The work introduces an innovative DMPC algorithm, enhanced with on-demand collision avoidance and event-triggered replanning mechanisms. The goal is to compute collision-free, goal-oriented paths for a swarm of robots, focusing on quadrotors as a testbed. Prior methods such as Sequential Convex Programming and Buffered Voronoi Cells (BVC) are benchmarked against the proposed solution, highlighting the significant improvements in travel time and success rates in dense environments.

The approach differentiates itself by achieving real-time collision avoidance without the conservativeness associated with BVC constraints, leveraging a more dynamic prediction-based method of trajectory adjustment. The concept of on-demand collision avoidance facilitates efficient space utilization by only applying restrictive measures when necessary, thus allowing agile navigation and better maneuverability in high-density operations.

Numerical Results and Claims

The paper reports that, compared to BVC, the proposed collision avoidance method reduces travel time by an average of 50%. Furthermore, the success rate for high-density agent scenarios is noted to exceed 90%, substantiating the efficacy of the DMPC formulation for complex multi-agent settings.

The real-time implementation is validated through simulation with up to 30 agents in a confined 18~$\text{m}^3$ arena, showcasing robust performance in collision avoidance and trajectory planning. Additionally, an empirical demonstration involving 20 quadrotors further reinforces the feasibility of the algorithm in practical settings.

Implications

On the practical side, this research advances the capabilities of multi-robot systems in dynamic environments, which are crucial for applications in logistics, surveillance, and autonomous navigation in urban settings. Theoretical implications revolve around the integration of predictive control and optimization strategies in distributed systems, providing groundwork for further exploration into improving real-time computation and decision-making for larger swarms or more complex environments.

Future Developments

The paper opens avenues for refining the communication model across agents to handle packet loss and delays, enhancing the robustness further when scaling the system. Additionally, investigating hybrid approaches combining discrete planners could optimize transition tasks involving non-convex obstacles, addressing potential deadlocks or local minima scenarios.

Overall, the research contributes a substantial enhancement to path planning methodologies within multi-agent systems, promising accelerated operation, heightened success probabilities, and adaptable real-time control mechanisms necessary for the growing complexity of autonomous multi-robot applications.