Persistent Intelligence, Surveillance, and Reconnaissance Using Multiple Autonomous Vehicles With Asynchronous Route Updates
Cameron K. Peterson, David W. Casbeer, Satyanarayana G. Manyam, Steven Rasmussen
Abstract
Providing persistent intelligence, reconnaissance, and surveillance of targets is a challenging, but important task when time-critical information is required. In this letter, we provide a decentralized routing algorithm for coordinating multiple autonomous vehicles as they visit a discrete set of pre-defined targets with weighted revisit priorities. The algorithm utilizes a block coordinate ascent algorithm combined with a Monte Carlo tree search to tractably decide each vehicle's route. The result is a non-myopic algorithm for multiple vehicles that is decentralized, computationally tractable, and allows for target prioritization. Guarantees are provided that all targets will have finite revisit times and that the block coordinate ascent algorithm will converge to a block optimal solution. Numerical simulations illustrate the utility of this method by showing that the results are comparable to those of a centralized exhaustive search and that they degrade gracefully with limited communication and scale under increasing numbers of targets and vehicles.
BibTeX
@inproceedings{ral2020_persistentintell,
title = {Persistent Intelligence, Surveillance, and Reconnaissance Using Multiple Autonomous Vehicles With Asynchronous Route Updates},
author = {Cameron K. Peterson and David W. Casbeer and Satyanarayana G. Manyam and Steven Rasmussen},
booktitle = {RA-L 2020},
year = {2020}
}