Attrition-Aware Adaptation for Multi-Agent Patrolling

TL;DR

The paper tackles multi-agent patrolling under agent attrition with minimal communication by introducing the Adaptive Heuristic-based Patrolling Algorithm (AHPA), which uses Voronoi partitioning and a two-stage decomposition to achieve guaranteed performance. It provides centralized and distributed mathematical formulations, along with theoretical performance bounds, and demonstrates robustness to attrition through a neighbor-only adaptation mechanism. AHPA combines a Voronoi-based node assignment with a nearest-neighbor visitation order to produce scalable, low-communication patrol plans, and its effectiveness is validated through both ROS 2 simulations and physical robot experiments. The work advances practical patrolling in disturbed environments by offering provable guarantees, reduced communication overhead, and efficient adaptation when agents drop out.

Abstract

Multi-agent patrolling is a key problem in a variety of domains such as intrusion detection, area surveillance, and policing which involves repeated visits by a group of agents to specified points in an environment. While the problem is well-studied, most works do not provide performance guarantees and either do not consider agent attrition or impose significant communication requirements to enable adaptation. In this work, we present the Adaptive Heuristic-based Patrolling Algorithm, which is capable of adaptation to agent loss using minimal communication by taking advantage of Voronoi partitioning, and which meets guaranteed performance bounds. Additionally, we provide new centralized and distributed mathematical programming formulations of the patrolling problem, analyze the properties of Voronoi partitioning, and finally, show the value of our adaptive heuristic algorithm by comparison with various benchmark algorithms using physical robots and simulation based on the Robot Operating System (ROS) 2.

Figures (5)

• Figure 1: Discontinuous weighted Voronoi partition based on heterogeneous agent speed, illustrating the dilemma described in \ref{['sec:mw-voronoi']}. Dots represent agents and colored boxes represent partitions assigned to the agent of the same color. The right-most agent (purple) has significantly higher speed than the other two, resulting in unexpected partitioning results.
• Figure 2: We use the MAS framework, Grex, for straightforward experimentation in both simulated environments and with physical robots. At top-left are the TurtleBot3 robots which we use in physical experiments. At top-right, simulated robots patrol the "Cumberland" environment from portugalDistributedMultirobotPatrol2013.
• Figure 3: Performance of the algorithms over time in the simulated Cumberland environment originally used in portugalDistributedMultirobotPatrol2013. At left, results with no agent attrition. Note the excellent comparative performance of AHPA (in blue). At right, with two instances of agent attrition. Note the cumulative message difference between AHPA and the other algorithms.
• Figure 4: Performance of AHPA and selected benchmark algorithms in real experiments. Note the stability of AHPA throughout. The DTAP and DTAG algorithms are not shown, as their performance was so poor that the graph became difficult to read.
• Figure 5: Performance of AHPA compared to the optimal performance. Note that AHPA remains within the performance bound described in \ref{['sec:bound-perf']}.

Theorems & Definitions (3)

• Definition 1: Neighboring Agent
• Theorem 1
• proof