Rolling Horizon Coverage Control with Collaborative Autonomous Agents

TL;DR

The paper tackles non-myopic coverage planning for a team of UAVs observing points on a 3D object's surface. It introduces a distributed rolling-horizon model predictive controller (P1) that jointly optimizes motion and camera inputs over a horizon $T$, incorporating light-path propagation constraints to infer visibility. Nonlinear visibility constraints are transformed into binary logic and embedded in a mixed-integer program, enabling scalable multi-agent planning with minimal work duplication. Evaluation includes simulations and real-world 3D architectural inspections, showing favorable computational scalability and effective coverage with partial or full cooperation. The results demonstrate that the distributed approach maintains performance while reducing planning time versus a centralized baseline.

Abstract

This work proposes a coverage controller that enables an aerial team of distributed autonomous agents to collaboratively generate non-myopic coverage plans over a rolling finite horizon, aiming to cover specific points on the surface area of a 3D object of interest. The collaborative coverage problem, formulated, as a distributed model predictive control problem, optimizes the agents' motion and camera control inputs, while considering inter-agent constraints aiming at reducing work redundancy. The proposed coverage controller integrates constraints based on light-path propagation techniques to predict the parts of the object's surface that are visible with regard to the agents' future anticipated states. This work also demonstrates how complex, non-linear visibility assessment constraints can be converted into logical expressions that are embedded as binary constraints into a mixed-integer optimization framework. The proposed approach has been demonstrated through simulations and practical applications for inspecting buildings with unmanned aerial vehicles (UAVs).

Figures (5)

• Figure 1: The figure illustrates: (a) the agent sensing model, (b) the triangulated surface area of the object of interest, and the points $\boldsymbol{p}_i$ that need to be covered, (c) the need for incorporating light-path propagation to asses the coverage of points i.e., although both points (marked with $\ast$, and $\times$) reside inside the agent's FOV, only point $\ast$ is visible as shown, this aspect is further discussed in Sec. \ref{['ssec:light_path_constraints']}.
• Figure 2: An illustrative example of the proposed collaborative 3D coverage controller, involving three agents indicated by green, red, and blue colors. (a) and (b) Predicted plans at time steps 1 and 6, respectively. (c) Allocation of points to the agents, and (d) the time step at which each point was covered.
• Figure 3: Coverage of the Marina Bay Sands Hotel in Singapore with 5 collaborative UAV agents. (a) Points of interest for coverage shown as $\circ$, (b)(c) Final coverage trajectories, (d) Allocation of points to agents, and (e) Timing of point coverage during the mission.
• Figure 4: Performance evaluation of the proposed approach. (a)(b) Computational complexity of centralized (PapaioannouCDC2023) and distributed (proposed approach) coverage planning with respect to the number of agents and the length of the planning horizon. (c) Performance comparison with competing approaches.
• Figure 5: Real-world experimental evaluation of the proposed collaborative 3D coverage controller: (a) Building for coverage, (b) Generated collaborative coverage plans with 3 UAV agents, (c) Real-time monitoring of mission progress, (d) DJI Mavic drones used in the experiment, (e) Telemetry data.