[Paper Review] Centralization vs. decentralization in multi-robot sweep coverage with ground robots and UAVs

Aryo Jamshidpey, Mostafa Wahby+4 more|arXiv (Cornell University)|Aug 13, 2024

Optimization and Search Problems4 citations

TL;DR

This paper investigates the trade-offs between centralized and decentralized control in multi-robot coverage using ground robots supervised by UAVs. It compares four control approaches—decentralized, hybrid, centralized, and predetermined—using ARGoS simulations, finding that hybrid control achieves near-centralized performance with strong scalability and fault tolerance, while centralized methods outperform decentralized ones in obstacle-rich environments despite higher energy costs.

ABSTRACT

In swarm robotics, decentralized control is often proposed as a more scalable and fault-tolerant alternative to centralized control. However, centralized behaviors are often faster and more efficient than their decentralized counterparts. In any given application, the goals and constraints of the task being solved should guide the choice to use centralized control, decentralized control, or a combination of the two. Currently, the exact trade-offs that exist between centralization and decentralization are not well defined. In this paper, we compare the performance of centralization and decentralization in the example task of sweep coverage, across five different types of multi-robot control structures: random walk, decentralized with beacons, hybrid formation control using self-organizing hierarchy, centralized formation control, and predetermined. In all five approaches, the coverage task is completed by a group of ground robots. In each approach, except for the random walk, the ground robots are assisted by UAVs, acting as supervisors or beacons. We compare the approaches in terms of three performance metrics for which centralized approaches are expected to have an advantage -- coverage completeness, coverage uniformity, and sweep completion time -- and two metrics for which decentralized approaches are expected to have an advantage -- scalability (4, 8, or 16 ground robots) and fault tolerance (0%, 25%, 50%, or 75% ground robot failure).

Motivation & Objective

• To systematically evaluate the trade-offs between centralization and decentralization in multi-robot coverage tasks.
• To assess how UAV supervision affects performance, scalability, and fault tolerance in ground robot coverage systems.
• To compare four control strategies—decentralized, hybrid, centralized, and predetermined—across metrics like coverage completeness, uniformity, energy use, and resilience.
• To determine under what environmental and operational conditions centralized or decentralized control is more effective.
• To evaluate the impact of UAV energy constraints and robot energy efficiency on system-wide performance and feasibility.

Proposed method

• Simulates multi-robot coverage tasks using the ARGoS physics-based simulation environment for large outdoor environments.
• Uses ground robots for actual coverage, while UAVs act as mobile supervisors providing global sensing and centralized coordination.
• Implements four control strategies: fully decentralized (reactive), hybrid (MNS formation with UAV coordination), centralized (global planning), and predetermined (fixed paths).
• Controls are compared by varying UAVs' access to global information and coordination authority, while ground robot capabilities remain unchanged.
• Measures performance using coverage completeness (fraction of area explored) and coverage uniformity (spatial distribution of exploration).
• Analyzes energy consumption based on real-world robot specifications, particularly UAV flight time and ground robot energy use.

Experimental results

Research questions

• RQ1 How does increasing centralization via UAV supervision affect coverage completeness and uniformity in multi-robot systems?
• RQ2 Does decentralized control outperform centralized control in environments with randomized obstacles, as commonly assumed?
• RQ3 To what extent does hybrid control achieve performance close to centralized control while maintaining scalability and fault tolerance?
• RQ4 How do energy consumption differences between UAVs and ground robots impact the feasibility and duration of centralized control strategies?
• RQ5 What role does autonomous recharging play in achieving high coverage completeness in large outdoor environments?

Key findings

• The predetermined control approach achieves the highest coverage completeness and uniformity, confirming the performance advantage of full centralization.
• The hybrid control approach performs significantly closer to centralized control than expected, achieving over 90% of the completeness of the best centralized method.
• Despite expectations, fully decentralized control underperforms in obstacle-rich environments, with centralized and hybrid approaches showing superior resilience to randomized obstacles.
• Hybrid control maintains strong scalability and fault tolerance comparable to fully decentralized control, even when individual robots fail.
• UAVs’ limited flight time is a more critical constraint than their energy efficiency, making long-duration centralized operations infeasible without recharging.
• Inefficient ground robots pose a greater performance bottleneck in centralized systems than inefficient UAVs, highlighting the need for energy-aware robot selection and autonomous recharging in large-scale deployments.