RF-Modulated Adaptive Communication Improves Multi-Agent Robotic Exploration

TL;DR

The paper addresses reliable coordination for multi-agent robotic exploration under limited communication by introducing Adaptive-RF Transmission (ART), a planning algorithm that selects transmission locations based on signal strength and payload size to minimize backtracking. An extension, ART-SST, imposes minimum signal thresholds to ensure high-fidelity data delivery. Through extensive simulations in three cave-inspired environments, ART consistently reduces traversal distance and exploration time, outperforming baseline strategies such as full rendezvous and simple signal-threshold policies, with gains up to around 58% in path length and 52% in exploration time. The work demonstrates that payload-aware, communication-guided planning substantially improves coverage efficiency in complex, communication-constrained settings and offers a foundation for future planetary exploration and search-and-rescue missions.

Abstract

Reliable coordination and efficient communication are critical challenges for multi-agent robotic exploration of environments where communication is limited. This work introduces Adaptive-RF Transmission (ART), a novel communication-aware planning algorithm that dynamically modulates transmission location based on signal strength and data payload size, enabling heterogeneous robot teams to share information efficiently without unnecessary backtracking. We further explore an extension to this approach called ART-SST, which enforces signal strength thresholds for high-fidelity data delivery. Through over 480 simulations across three cave-inspired environments, ART consistently outperforms existing strategies, including full rendezvous and minimum-signal heuristic approaches, achieving up to a 58% reduction in distance traveled and up to 52% faster exploration times compared to baseline methods. These results demonstrate that adaptive, payload-aware communication significantly improves coverage efficiency and mission speed in complex, communication-constrained environments, offering a promising foundation for future planetary exploration and search-and-rescue missions.

Figures (6)

• Figure 1: The Scout-Specialist robot team starts co-located in a simulated cave environment. The Scout flies ahead to scan the area, taking signal strength estimates to assess the communication link strength between the two robots. Shown here is the top down view of the signal strength estimations generated by the Scout vehicle as it explores the environment. A small opening provides an opportunity for Scout to transmit to the Specialist with higher bandwidth signal, even though it cannot traverse through the opening. From the samples the Scout has taken, a communication-aware motion planning algorithm determines the best trade-off between signal strength for a data payload size and the time cost of navigating to different locations before transmitting.
• Figure 2: We collect signal strength measurements between two hardware platforms in a narrow corridor for baseline signal strength ranges in a real-world deployment. In line of sight settings, the signal strength remains strong for 40 meters. However, as soon as the Scout agent turns a corner and enters a perpendicular hallway, the signal strength rapidly declines, eventually resulting in a lost connection (depicted with a red x). The observed signal drop-off highlights the importance of adaptive transmission strategies that account for payload size and connectivity when deploying coordinated multi-agent systems.
• Figure 3: The system uses a custom state machine, frontier finder, and adaptive-RF module to evaluate the communication-aware motion planning algorithms described in Section \ref{['sec:evaluation']}. The custom software components integrate with standard ROS2 packages: NAV2 and slam_toolbox. Each robot runs the same software stack and communication takes place over ROS2 message exchange when the agents are within transmission range of one another.
• Figure 4: Top left: visible skylight at Wind Cave in Logan, Utah, USA Carney2020_WindCaveGeoSights. Top right: Permafrost Research Center in Fairbanks, Alaska, USA USACE2025_PermafrostTunnelPhoto. These real-world environments inspire the design of simulated testing domains used in this work. The window environment (bottom left) models the skylight geometry while the Y-junction environment (bottom middle) reflects the branching structure of the permafrost tunnel. A third environment (bottom right), a long and narrow tunnel, forces agents out of communication range, enabling evaluation of algorithm performance under communication-denied conditions.
• Figure 5: Comparison of mean total path lengths for all algorithms across all payload sizes (0–3) in the Long Tunnel (LT) environment. ART consistently achieves the shortest paths across all payload sizes, indicating more efficient exploration. ART-SST, MSSC, and FRC see substantially longer paths, with FRC performing the worst overall. Error bars represent one standard deviation.