FG-PE: Factor-graph Approach for Multi-robot Pursuit-Evasion

TL;DR

FG-PE introduces a factor-graph formulation for coordinating multiple pursuers against a single evader in a 2D environment. By unifying evader estimation and pursuer planning under MAP inference and leveraging a Levenberg–Marquardt optimizer within the GTSAM framework, FG-PE explicitly handles uncertainty and communication dropouts while scaling to more agents and obstacles. The approach yields substantial reductions in capture time and travel distance compared to baselines, remains robust under reduced measurement frequency and dropped messages, and is validated through both simulated and real-world TurtleBot experiments. This work provides a principled, scalable, and uncertainty-aware tool for cooperative pursuit-evasion with strong practical implications for multi-robot surveillance and search-and-rescue tasks.

Abstract

With the increasing use of robots in daily life, there is a growing need to provide robust collaboration protocols for robots to tackle more complicated and dynamic problems effectively. This paper presents a novel, factor graph-based approach to address the pursuit-evasion problem, enabling accurate estimation, planning, and tracking of an evader by multiple pursuers working together. It is assumed that there are multiple pursuers and only one evader in this scenario. The proposed method significantly improves the accuracy of evader estimation and tracking, allowing pursuers to capture the evader in the shortest possible time and distance compared to existing techniques. In addition to these primary objectives, the proposed approach effectively minimizes uncertainty while remaining robust, even when communication issues lead to some messages being dropped or lost. Through a series of comprehensive experiments, this paper demonstrates that the proposed algorithm consistently outperforms traditional pursuit-evasion methods across several key performance metrics, such as the time required to capture the evader and the average distance traveled by the pursuers. Additionally, the proposed method is tested in real-world hardware experiments, further validating its effectiveness and applicability.

Figures (17)

• Figure 1: The factor graph representation for the PE problem. Green nodes represent the variables associated with the poses of the pursuers. Red nodes are poses of evaders. Squares represent the constraints between the variables, including motion models (e.g. $f_1^{dp_1}$), measurements of the robots (e.g. $f_0^{mp_1}$) or landmarks (e.g. $\textcolor{black}{f_1^{mo_{11}}}$).
• Figure 2: The number of different factors at each time step is shown in the figure above. In this figure, $f^{op}$ and $f^{mo}$ are increasing at the same rate, while $f^{dp}$, $f^{mp}$, and $\textcolor{black}{f^{movp}}$ are increasing at the same speed.
• Figure 3: A simplified factor graph with a tree structure is presented. In this scenario, there is only one measurement from each pursuer and evader.
• Figure 4: Time to capture the evader for four pursuers in 1000 iterations to find the optimal weights for our problem.
• Figure 5: Correlation analysis between parameters and time to capture the evader (See Table \ref{['vars_avoidance_extended']} for parameters).