Control of Artificial Swarms with DDDAS R Ryan McCune & Greg Madey Department of Computer Science & Engineering, University of Notre Dame 2014 InternaConal Conference on ComputaConal Science Dynamic DataDriven ApplicaCon Systems (DDDAS) Track June 10, 2014 Cairns, Australia
18
Embed
Control of Artificial Swarms with DDDAS · Control of Artificial Swarms with DDDAS RRyan"McCune"&"Greg" Madey" " Departmentof"Computer"Science"&"Engineering,"University"of"Notre"Dame"
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Control of Artificial Swarms with DDDAS
R Ryan McCune & Greg Madey
Department of Computer Science & Engineering, University of Notre Dame 2014 InternaConal Conference on ComputaConal Science Dynamic Data-‐Driven ApplicaCon Systems (DDDAS) Track
June 10, 2014 Cairns, Australia
Overview • Mo#va#ng problem – UAVs • Solu#on – Swarm Intelligence • Approach – DDDAS • Framework for swarm control with DDDAS • Applica#on example – Swarm intelligent ant foraging – UAV clustering
1
Operator Overload • Unmanned Aerial Vehicles (UAVs) – No on-‐board pilot – Intelligence, Surveillance, and Reconnaissance (ISR) missions – Becoming smaller and cheaper with increased capabili#es
• How to efficiently operate many UAVs?
2
Swarm Intelligence • Biologically inspired – Ant colonies, flocks of birds, fish
• Emergent phenomena – Simple, local behavior of agents – Complex, global behavior of system
• Self-‐organiza#on – Decentralized
• Emergent behavior solves problems
3
DDDAS • Entails: – Dynamic incorpora#on of addi#onal data into execu#ng simula#on
Framework for Swarm Control with DDDAS • Swarm Applica#on Architecture – One (or few) agent parameters – Swarm performance reported by single sta#s#c
• Feedback Control Loop – Calibrate with real-‐#me data – Sweep of parameters – Op#miza#on via simula#on
5
Swarm Application Architecture
6
DDDAS Feedback Control Loop
7
Example – General Ant Foraging
8
• Ants search for food to bring back to nest
• Randomly search environment • Deposit pheromones while searching – Likely to follow pheromones – Random Ac#on Probability (RAP)
• Shortest path emerges
RAP = ρ
1 – ρ Follow highest pheromone
ρ Random direc#on
Ant Foraging - An Implementation[1] • Ants deposit 2 pheromones – Green lead to home, deposit while foraging – Blue lead to food, deposit while returning home
9
[1] Panait, Liviu, and Sean Luke. "A pheromone-‐based u#lity model for collabora#ve foraging." Proceedings of the Third Interna#onal Joint Conference on Autonomous Agents and Mul#agent Systems-‐Volume 1. IEEE Computer Society, 2004.
• 1 ant hill – Sta#onary
• 1 food – Unlimited
• Many ants
10
[1] Panait, Liviu, and Sean Luke. "A pheromone-‐based u#lity model for collabora#ve foraging." Proceedings of the Third Interna#onal Joint Conference on Autonomous Agents and Mul#agent Systems-‐Volume 1. IEEE Computer Society, 2004.
Ant Hill
Food
Swarm Clustering • Adapted from ant foraging – Many food instead of 1 food – Many ant hills instead of 1 ant hill • Ant hills can move (right)
– Only 1 pheromone type, not 2 • Deposit when looking for food • Follow to return to ant hill • No pheromone leads to food • Once any food is found randomly, pheromone leads to nearest ant hill