170 likes | 440 Views
Simulation and Optimization of a Collaborative Multi-Robot Heterogeneous Firefighting System. Dimitrios Bakllas , Aaron Fineman and Tanay Shah. ME 593 Autonomous Multi-Robot Systems. 06/04/2011. Outline. Introduction Background Mapping Player-Stage Assumptions Robots Algorithm.
E N D
Simulation and Optimization of a Collaborative Multi-Robot Heterogeneous Firefighting System DimitriosBakllas, Aaron Fineman and TanayShah ME 593 Autonomous Multi-Robot Systems 06/04/2011
Outline Introduction Background Mapping Player-Stage Assumptions Robots Algorithm
Introduction Distributed Firefighting Locate Fires Put Out Fires Distributed Mapping Update Blueprint Track Fires
Introduction/Distributed Firefighting Heterogeneous swarm Scouting Bots Firefighting Bots http://www.toytokyo.com/shopping/index.php/page/product/product_id/5245 http://www.robot.uji.es/lab/plone/robots
Introduction/Distributed Mapping Simultaneous Localization and Mapping (SLAM) Distributed Mapping of the Environment Environment is dynamic Estimation of fire expansion
Mapping Standard 2D X-Y maps 4D maps! Trustworthiness of merged data Fire Stored as bitmaps #RRGGBB Provides information over time
Player-Stage • Robot abstraction software • Player • Runs on the robot • Provides a common interface to various robots • Runs over the network "All the world's a stage, And all the men and women merely players."- William Shakespeare, As You Like It
Stage • Stage • Robot Simulator
Player-Stage • Gazebo • 3D version of Stage http://playerstage.sourceforge.net/
Assumptions The firefighting robots will be limited in number They are heat resistant but impervious to fire and rubble Can move easily over terrain except rubble Firefighters are slow to due to weight of equipment Firefighters have infinite water All robots have infinite battery To simulate smoky environment communication range will be limited Firefighters have double communication range
Robots • Scouts have standard IR sensors. Range of 20-150 cm • They also have sonar sensors. Range of 0 to 645 cm with resolution of 2.5 cm • SICK LIDAR sensors with field view of 360 and operating range of 10 to 2000 cm
SLAM SLAM will be used to update the preloaded map within the known environment and at the same time keep track of the current location of each individual agent. foreach sensor in sensor types[] : foreachsensor in array[] : sensor matrix[m,n] = sensorProxy.scan[n] + normal(0, error) location = filter noise(sensor matrix[],time slice)
First Come First Serve (FCFS) With FCFS, fires will be attended to in the order they were reported. list fire locations[]: foreacht in time : foreachrobot in swarm[] : if(FOUND FIRE FLAG == True) : fire locations.add(robot.location, robot.sensor.temp) foreachfirefighter in swarm[] : if(fire locations.size > 0) goto(fire locations[1]) pop(fire locations[1])
Greedy Algorithm With the greedy algorithm, the room with the most reported fires will be attended to first. list rooms[] while(1) : foreachrobot in swarm[] : If(FOUND FIRE FLAG == True) : rooms[robot.getroom()] += 1; foreach firefighter in swarm[] : sorted rooms = rooms.sort(); if(sorted rooms[1] > 0) : goto(sorted rooms[1]) sorted rooms[1] = 0