Source-linked AI summary
Autonomous robotic bridging using distributed swarm control without inter-agent communication
Vishwaak C. Thamaraiselvan, Cody L. Lundberg, Michail Theofandis, Suhas Chelian, Nicholas R. Gans
TL;DR
SCARAB addresses autonomous robotic bridging for contested river crossings, where existing technologies do not account for near-peer threats. It combines distributed formation control and multi-modal sensing, demonstrating autonomous USV raft formation in physics-engine simulation.
Problem
The Army needs autonomous robotic bridging that supports dispersed river crossings and sustainment in contested environments while improving crew survivability and reducing logistics footprint.
Method
SCARAB combines distributed nonlinear formation control with multi-modal sensing and localization for USVs and docking targets, without requiring GPS, shared heading, or inter-agent communication.
Results
SCARAB demonstrated autonomous USV raft formation in simulation; the primary formation converged in 492 time steps, or 49.2 seconds of simulation time.
Takeaways & Limitations
Floating bays can assemble autonomously using distributed swarm control and multi-modal perception, with optional RF or acoustic communication.
Takeaways & Limitations
Further work is needed for terrain obstacles, adverse vision conditions, and formations containing damaged, disabled, or non-responsive units.
Abstract
from arXiv · showhide
We describe SCARAB--Swarm-Capable Autonomous Robotic Aquatic Bridging. Using distributed swarm control and multi-model sensing of agents and docking targets, agents can localize themselves, join into formations and proceed to desired target locations. Our technologies would eventually allow the Army to perform unpredictable, dispersed river crossings, enhance crew survivability, and minimize the logistics footprint compared to the current Improved Ribbon Bridge. Our methods operate without GPS or RF communications, though these can be used in non-contested environments (e.g., civilian disaster relief for flooding, etc.). We demonstrate our system via physics-engine-based simulation of several agents using unmanned surface vehicles (USVs) in the presence of currents and wind.
1. INTRODUCTION
SCARAB targets autonomous robotic bridging for contested river crossings by combining distributed formation control with multi-modal sensing. The system is designed to operate without GPS or inter-agent communication, while optionally incorporating them for civilian applications.
- SCARAB addresses autonomous robotic bridging for unpredictable, dispersed river crossings, crew survivability, and reduced logistics footprint.
- Autonomous bays form a 2x3 rafting formation to carry large loads without GPS or inter-agent communication.
- The system combines distributed nonlinear formation control with multi-modal sensing and localization for USVs and docking targets.
- GPS and RF communications remain optional for non-contested settings such as civilian flood-disaster relief.
2. BACKGROUND
The paper develops a distributed formation-control approach based on barycentric coordinates to avoid global sensing, common headings, and inter-agent communication. It extends this framework to heterogeneous vehicle dynamics and combines it with sensing, localization, and collision avoidance.
- Many formation-control methods require GPS or peer communication, while distance- or bearing-based methods can converge to incorrect local formations.
- Barycentric coordinates enable distributed formation control without global position measurements, common heading direction, or centralized coordination.
- The proposed strategy is distributed, provably convergent, and robust across USV, UAV, and UGV dynamics.
- The controller remains stable under significant disturbances and input saturation and includes TRL 4+ distributed collision avoidance with stability guarantees.
- Barycentric-coordinate control regulates each agent relative to neighboring agents, supporting prescribed formations under heterogeneous dynamics.
- Prior demonstrations included formation control for wheeled ground vehicles and aerial vehicles, with swarm formations of up to nine units.
3. APPROACH
SCARAB combines VRX-based USV simulation, multimodal localization, and distributed feedback control to support formation behavior without requiring global sensing or inter-agent communication. The approach fuses camera, LIDAR, GPS, and IMU information while evaluating formation convergence with explicit error and time metrics.
- Simulation software and sensing: VRX provides a Gazebo-based simulation environment with modeled hydrodynamics, currents, waves, thrusters, and multiple USV sensors.The simulator includes cameras, GPS, IMU, and 3D LIDAR through the WAM-V configuration.
- Simulation software and sensing: The simulation adapts the Hydronalix Reckless frame by importing its 3D mesh and modifying vehicle dynamics, propulsion, steering gains, and steering limits.Multiple Reckless-based USVs are spawned in the simulator.
- Simulation software and sensing: AprilTag computer vision localizes GPS-denied docking targets, while LIDAR provides a baseline for assessing camera-based localization accuracy.The AprilTag cube is used as a static target, with future work aimed at localizing arbitrary visual docking targets.
- Simulation software and sensing: Camera-detected AprilTag centroids and LIDAR point clouds agree on the cube location, diverging by less than 5%.The centroids lie on the cube surface as represented by LIDAR.
- Localization fusion: Fused USV localization combines GPS and reversed AprilTag estimates using a short buffer and an exponentially weighted moving average.The weighting balances quick state updates with reduced sensitivity to outlier position disturbances.
- Vehicle control model: The USV control architecture separates heading and speed regulation: formation commands produce desired heading and speed, which become steering and velocity inputs.Heading uses IMU orientation, while speed uses GPS-derived position and velocity information; the vehicle is actuated by two podded thrusters.
- Swarm formation control: The formation controller uses local relative positions from onboard vision and includes collision avoidance, avoiding dependence on GPS, shared orientation, or inter-agent communication.The framework is extended to USVs and validated for formation convergence under dynamic uncertainty and environmental disturbances.
- Swarm formation control: Formation convergence is measured with convergence time et and mean formation error ef, with et requiring ef < ϵ for 10 consecutive simulation steps.The proof-of-concept tolerance is ϵ = 1 m and may be reduced in future work.
4. EXPERIMENTS
The experiments simulate six USVs forming either a 2x3 raft or an X configuration with 1 m inter-agent spacing. Both configurations converge under the distributed controller, with the 2x3 formation converging in 49.2 seconds of simulation time and the X formation in 33 seconds.
- 4. EXPERIMENTS: Six USVs were simulated in Gazebo using the Hydronalix Reckless platform, evaluating 2x3 raft and X-shaped target formations.The formations used 1 m inter-agent spacing.
- 4. EXPERIMENTS: By step 482, all agents satisfied the 2x3 formation convergence condition, corresponding to et = 492 time steps and 49.2 seconds of simulation time.Each simulation time step represented 0.1 seconds.
- 4. EXPERIMENTS: The X configuration converged in et = 332, corresponding to 33 seconds of simulation time and 3 minutes 59 seconds of wall-clock time.The reported wall-clock duration reflects the simulation real-time factor.
- 4. EXPERIMENTS: Five bays initialized at random positions converged to an X-shaped formation.Figures 9 and 10 present the progression and error behavior for this alternate swarm-control result.
- 4. EXPERIMENTS: Figure 8 plots mean formation error ef and each agent’s position error against simulation time step, with a dashed convergence threshold.The plot compares per-agent errors with the mean formation error for the 2x3 formation.
5. CONCLUSION
The conclusion presents SCARAB as an autonomous aquatic-bridging system in which floating bays assemble through distributed swarm control and multimodal perception. It identifies robustness extensions involving obstacles, adverse visual conditions, and damaged or non-responsive units.
- 5. CONCLUSION: SCARAB enables floating bays to assemble autonomously for robotic aquatic bridging using distributed swarm control and multimodal perception.RF or acoustic-modem communication may optionally be incorporated.
- 5. CONCLUSION: The authors identify obstacle avoidance, robust vision in smoke, fog, or low light, and formation control with damaged or non-responsive units as extensions.The passage cites potential-fields obstacle handling, infrared target detection, and replacement-bay strategies as related capabilities or options.
7. CONTACT INFORMATION
The contact information identifies Nicholas R. Gans as a professor and principal scientist at the University of Texas at Arlington.
- 7. CONTACT INFORMATION: Nicholas R. Gans, PhD, is listed as Professor of Computer Science and Engineering and Principal Scientist at the University of Texas at Arlington.