Precise Formation of Multi-Robot Systems

chestpeeverIA et Robotique

13 nov. 2013 (il y a 8 années et 3 mois)

363 vue(s)

Precise Formation of Multi-Robot
Christopher M. Gifford and Arvin Agah
Center for Remote Sensing of Ice Sheets (CReSIS)
University of Kansas
April 17, 2007
•Results and Analysis
•High-precision sensor placement applications
–Unstructured, remote environments are complex
–High precision may not be easily attainable
–Sparsely spaced sensors leave few viable options
–Example: Geophysical (seismic) surveying
•Mobile robot team
–Can provide a higher level of precision
–Can remove or greatly reduce the human footprint
–Systems of systems of systems
Seismic Surveying
•Seismic sensors (geophones)
–Ground vibrations mapped to analog signals
–2D or 3D “images”of the subsurface
•Precision grid spacing, shape, and orientation
–Centimeters error of perfect grid or line desired
–Linear or rectangular grid, varying or equal spacing
•Robot team represents entire seismic array
–Mobile robot represents a single geophone in the grid
•High-precision, shape-based grid formation
scheme for multi-robot systems
–Can be used in dense or sparse applications
–Incremental deployment algorithm
–Form desired shape using GPS coordinates
•Relationships between traveling speed,
formation time, energy usage, grid shape,
and positioning error
•Sonar arrays not feasible (imprecise)
•One or more cameras can track relative
distances, but are limited in precision
•Beacons not always readily present
•Laser range-finders (LRFs) decrease in
accuracy as distance increases
•Global positioning systems (GPS) provide
precision in all environments, for all spacings
•Positioning error
–Terrain composition and state (e.g., slippage on snow)
–Obstacles and hazards (e.g., crevasses)
–Weather and lighting
–Error inherent in GPS and mobile devices
•Simulation study using Webots
–Each mobile robot has its own GPS receiver
–Each can communicate with a larger, main robot
–Each is small, has four wheels, and onboard battery
•Small team of
25 simulated
mobile robots
getting off trailer
•Robot team
initial position
for all simulation
Approach: Grid Shape and Spacing
5x5 Square Grid, Larger Spacing
8x3 Rectangular Grid, Small Spacing
Approach: Example Simulation
Experiments: Setup
•Simulated 12 grid formations
1. Squares: 5x5, 4x4, 3x3, 2x2
2. Rectangles: 8x3, 6x4, 4x3, 3x2
3. Lines: 4x1, 8x1, 16x1, 24x1
•Varied 3 levels of speed
1. Slow (25% full speed)
2. Normal (50% full speed)
3. Fast (75% full speed)
•Each combination run 4 times = 144 simulations
Experiments: Assumptions
•All GPS receivers are highly precise (centimeters)
•All robots turn at same speed (5% of full speed)
•All driving speeds are constant (non-changing)
•Robots can stop immediately (instantly)
•Battery usage divided into 99% motors, 1% CPU
•No wheel slippage or wind force
•Simulated GPS error is random
•Robots initially 0.5 meters apart in Xand Ydirections
•Final grid will have 10 meter uniform spacing
•Manhattan Distance used for positioning error
Results: Formation Time
Results: Positioning Precision
•Overall precision decreased as more robots involved
•Square-like shapes with high speeds had most error
•Confirms speed greatly effects positioning error
Precision (Error) Versus Speed
Square Grids
•Larger: less effected Slow to Normal
•Smaller: less effected Normal to Fast Rectangular Grids
•Square-like 6x4 grid had most error
•Smaller: less error Linear Grids
•Longer: less effected Slow to Normal
•Shorter: less effected Normal to Fast
•Error comparable for all lengths
Results: Double Shape Dimensions
•High-precision, incremental grid formation approach
using a team of mobile robots
•Tradeoff: formation time, precision, and collision risk
•Results show several patterns that effect grid precision
–The longer robots must travel, more error was introduced
–More robots caused more overall grid positioning error
–Higher precision could be attained by driving at slower speeds
–More square shapes exhibit more positioning error
•Varying grid spacing could also be performed to learn
more about how it effects formation variables