Autonomous Decision In Multi Robot Systems
Abstract
This research evaluates potential auction algorithm approaches to a multi-robot area search problem and uses the Naval Postgraduate School Advanced Robotic System Engineering Laboratorys multi-UAV system to implement, test, and evaluate selected exemplars. Ultimately, for multi-robot systems to achieve useful objectives autonomously, they need to reliably analyze objectives and assign supporting tasks to individual vehicles. The market-based approaches analyzed in this research provide anintuitive mechanism for robust realization of this capability in highly dynamic and uncertain environments. We present our implementation, AuctionSearch, evaluate its design trade-offs, and influence agent bidding strategies based on per-robot speed and endurance. We test our implementation in simulation and in live-fly experiments acrossthree different search areas with system sizes ranging from three to 10 robots each. The future of warfare will include unmanned systems in many facets of operations and support. Furthermore, it is likely that human intervention and direct handling ofautonomous systems actions will be replaced by human supervision of autonomously developed courses of action on the battlefield. For multi-robot systems to have the capacity to develop and execute complex courses of action, they must be capable of linking complex tasks together. Our research and testing demonstrate that auction algorithms are well suited for autonomous decision.
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 01, 2018
- Accession Number
- AD1069610
Entities
People
- Matthew S. Hopchak
Organizations
- Naval Postgraduate School