Don’t Crowd Me Summary of and comments on Brogan and Hodgins’ Group Behaviors for Systems with...

Post on 18-Jan-2018

217 views 0 download

description

Significant Dynamics Significant Dynamics: the motion of an individual is significantly affected by its dynamic properties.Significant Dynamics: the motion of an individual is significantly affected by its dynamic properties. –Acceleration –Velocity –Turning radius

Transcript of Don’t Crowd Me Summary of and comments on Brogan and Hodgins’ Group Behaviors for Systems with...

Don’t Crowd MeDon’t Crowd MeSummary of and comments on Brogan and Hodgins’Summary of and comments on Brogan and Hodgins’

Group Behaviors for Systems with Significant DynamicsGroup Behaviors for Systems with Significant Dynamics

Cailin K. AndrussCailin K. AndrussVirginia Commonwealth UniversityVirginia Commonwealth University

NSF BBSI Program 2005 - 2006NSF BBSI Program 2005 - 2006

IntroductionIntroduction

• To move as a group, animals must To move as a group, animals must – Maintain close proximityMaintain close proximity– Avoid collisionsAvoid collisions– Compensate for significant dynamicsCompensate for significant dynamics

Significant DynamicsSignificant Dynamics

• Significant Dynamics: the motion of Significant Dynamics: the motion of an individual is significantly affected an individual is significantly affected by its dynamic properties.by its dynamic properties.– AccelerationAcceleration– VelocityVelocity– Turning radiusTurning radius

The GoalThe Goal

• An algorithm An algorithm – Natural looking low-level behaviorsNatural looking low-level behaviors

• Walking, running, bicycling, etc.Walking, running, bicycling, etc.– Realistic interaction with an Realistic interaction with an

unpredictable environmentunpredictable environment• Obstacle avoidance, grouping, rough Obstacle avoidance, grouping, rough

terrain locomotion, etc.terrain locomotion, etc.

• Can be applied to physical robots Can be applied to physical robots performing useful tasksperforming useful tasks

The AlgorithmThe Algorithm

PerceptionPerception

PlacementPlacement Obstacle Obstacle

AvoidanceAvoidance

The AgentsThe Agents

The AgentsThe Agents

• One-legged RobotsOne-legged Robots

The AgentsThe Agents

• One-legged RobotsOne-legged Robots

• Rigid-body model of Rigid-body model of human riding bicyclehuman riding bicycle

The AgentsThe Agents

• One-legged RobotsOne-legged Robots

• Rigid-body model of Rigid-body model of human riding bicyclehuman riding bicycle

• Point-mass systemPoint-mass system

ResultsResults

ResultsResults

GroupingGrouping

ResultsResults

TurningTurning

(Click here for movie)

ResultsResults

Obstacle AvoidanceObstacle Avoidance

Results SummaryResults Summary• Point-masses: moved more tightly Point-masses: moved more tightly

under changes in velocity because of under changes in velocity because of the more exact control of velocity.the more exact control of velocity.

• Robots: more variability and motion Robots: more variability and motion within the group. Separation distance within the group. Separation distance was made larger to prevent collisions was made larger to prevent collisions between members. between members.

• Bicyclists: the control system was not Bicyclists: the control system was not as robust, and the they were not able as robust, and the they were not able to perform as well on the turning test.to perform as well on the turning test.

LimitationsLimitations

• Reflexive reactions to collisionsReflexive reactions to collisions

• Information about other individuals Information about other individuals is too accurateis too accurate

• Models are simplified Models are simplified

• Heterogeneous populationHeterogeneous population

ReferencesReferences

Brogan, D.C., Hodgins, J. K., 1997. Brogan, D.C., Hodgins, J. K., 1997. Group Group Behaviors for Systems with Significant Behaviors for Systems with Significant DynamicsDynamics. . Autonomous Robots Autonomous Robots 4, 4, p. 137-153.p. 137-153.