Coordinated Multi-Robot Navigation with Formation Adaptation

TL;DR

AFOR addresses the challenge of coordinated multi-robot navigation in environments where rigid formations fail, by integrating hierarchical learning with a spring-damper-based formation adaptation mechanism. The upper-level graph neural network enables team-level decision making and information sharing, while the lower-level reinforcement learning controller handles obstacle avoidance and individual navigation; the spring-damper component smooths formation changes and reduces oscillations. Trained with Proximal Policy Optimization, AFOR jointly optimizes both levels to maintain adaptable formations and efficient progress toward goals, demonstrated in Gazebo, Unity3D, and real robot experiments across circle, wedge, and line formations. The results show superior performance in formation integrity and navigation success compared to learning-free baselines, leader-follower schemes, and decoupled GNN methods, highlighting AFOR’s practical potential for scalable, robust multi-robot coordination in cluttered environments.

Abstract

Coordinated multi-robot navigation is an essential ability for a team of robots operating in diverse environments. Robot teams often need to maintain specific formations, such as wedge formations, to enhance visibility, positioning, and efficiency during fast movement. However, complex environments such as narrow corridors challenge rigid team formations, which makes effective formation control difficult in real-world environments. To address this challenge, we introduce a novel Adaptive Formation with Oscillation Reduction (AFOR) approach to improve coordinated multi-robot navigation. We develop AFOR under the theoretical framework of hierarchical learning and integrate a spring-damper model with hierarchical learning to enable both team coordination and individual robot control. At the upper level, a graph neural network facilitates formation adaptation and information sharing among the robots. At the lower level, reinforcement learning enables each robot to navigate and avoid obstacles while maintaining the formations. We conducted extensive experiments using Gazebo in the Robot Operating System (ROS), a high-fidelity Unity3D simulator with ROS, and real robot teams. Results demonstrate that AFOR enables smooth navigation with formation adaptation in complex scenarios and outperforms previous methods. More details of this work are provided on the project website: https://hcrlab.gitlab.io/project/afor.

Figures (7)

• Figure 1: A motivating scenario for coordinated multi-robot navigation with formation adaptation. The team of robots must dynamically adapt their desired formation to navigate through a narrow bridge while smoothly moving towards their goal positions.
• Figure 2: Overview of AFOR for coordinated multi-robot navigation with formation adaptation. AFOR's unified hierarchical learning model uses an upper-level GNN to share information and make team coordination decisions, while the lower-level RL learns individual robot control for navigation and obstacle avoidance. AFOR also integrates a spring-damper model with the hierarchical learning to enable formation adaptation and reduce oscillations.
• Figure 3: Qualitative results on coordinated navigation with robots maintaining circle, wedge and line formations in Gazebo simulations using ROS.
• Figure 4: The trajectories depict a team of robots navigating a narrow corridor while adaptively maintaining formations. Each robot's path is represented by a different color, with key timestamps marking formation transitions. The gray lines indicate obstacles.
• Figure 5: Qualitative results on formation adaptation for coordinated multi-robot navigation using real mobile robots that form circle, wedge and line formations.