LOMORO: Long-term Monitoring of Dynamic Targets with Minimum Robotic Fleet under Resource Constraints

TL;DR

The work tackles long-term monitoring of many dynamic targets with a resource-constrained robotic fleet. It introduces LOMORO, a hierarchical online coordination framework that couples search-based robot ordering, a time-varying node network with an Incremental Martin's Algorithm for task routing and charging, and online adaptation to unknown target dynamics and failures. The method guarantees bounded monitoring intervals and resource levels while minimizing the average fleet size, and demonstrates scalability to large road networks and high target counts. Extensive simulations across varying network topologies, speeds, charging rates, and monitoring requirements demonstrate improved scalability, robustness to failures, and efficient charging management, advancing online, resource-aware surveillance in unknown dynamic environments.

Abstract

Long-term monitoring of numerous dynamic targets can be tedious for a human operator and infeasible for a single robot, e.g., to monitor wild flocks, detect intruders, search and rescue. Fleets of autonomous robots can be effective by acting collaboratively and concurrently. However, the online coordination is challenging due to the unknown behaviors of the targets and the limited perception of each robot. Existing work often deploys all robots available without minimizing the fleet size, or neglects the constraints on their resources such as battery and memory. This work proposes an online coordination scheme called LOMORO for collaborative target monitoring, path routing and resource charging. It includes three core components: (I) the modeling of multi-robot task assignment problem under the constraints on resources and monitoring intervals; (II) the resource-aware task coordination algorithm iterates between the high-level assignment of dynamic targets and the low-level multi-objective routing via the Martin's algorithm; (III) the online adaptation algorithm in case of unpredictable target behaviors and robot failures. It ensures the explicitly upper-bounded monitoring intervals for all targets and the lower-bounded resource levels for all robots, while minimizing the average number of active robots. The proposed methods are validated extensively via large-scale simulations against several baselines, under different road networks, robot velocities, charging rates and monitoring intervals.

Figures (5)

• Figure 1: Illustration of the considered scenario. Top-Left and Top-Right: $3-4$ UAVs are actively monitoring $10$ targets (in red) within a road network, with their static (t=$12s$) and online (t=$381s$) plans. Middle Left: Batteries of 6 robots during the online execution of $400s$. Bottom-Left: Intervals from the last monitoring of 10 targets. Middle-Right: Average number of active robots, the time when replans take place, and their computation time. Bottom-Right: The robots responsible for each target during any consecutive replans. Each target has 2 rows, with the lower representing the target node and the upper the intersection target node respectively.
• Figure 2: Illustration of the proposed framework, which consists of three parts: Search-based optimization for robot ordering, multi-robot maximum-allowed Martin's algorithm (MAM) and online execution. The MAM consists of the time-varying node network (TVNN) and the incremental Martin's algorithm (IMA). The IMA mainly consists of incremental sub-TVNN and the Martin's Process.
• Figure 3: Top-Left and Top-Right: Trajectories of 8 robots actively monitoring 15 targets in the considered scenario at t=$84s$ and t=$347s$ respectively. Bottom: The action of 8 robots at each time during the simulation of $400s$.
• Figure 4: Illustration of scalability analysis results. The number of active robots with an increasing number of targets, concerning $\gamma_n/\beta_s$(a), $T_m$(b) and $C_n$(c). The computation time with an increasing number of targets, concerning $C_n$(d).
• Figure 5: Illustration of the generalization circumstances. (a): Previous plan result at $t=20s$ when a robot accidentally crashes. (b): New replan after the robot crashed. (c): Trajectories of 6 robots monitoring 15 targets that can dynamically enter and exit the road network over a period of $400s$