Behavior-Based Formation Control for Multi-robot Teams

Tucker Balch, and Ronald C. Arkin

Introduction Main idea: Behavior-based approach to robot

formation-keeping. The robots in this research have

different IDs: heterogeneous robot. The formation behaviors were implemented as

motor schema within AuRA, and as steering and speed behaviors within UGV demo II architecture.

Background Why formation:

Formation in nature:1. To detect predators2. To efficiently forage for food Grouping Multiple sensor

Formation in robot: • Robotic scouts also benefit by directing

their sensors in different areas

Craig Reynolds pioneering work: • A simple egocentric behavioral model

for flocking• The behavior consists of several

separate components: 1) collision avoidance 2) Velocity matching3) Flock centering

Tu and Terzopoulos:• Realistic simulated fish schooling.

Brogan and Hodgins:• System for realistically animating herds

of one-legged agents using dynamical models of robot motion.

Wang:• Strategy for robot formation.• Individual robots are given specific

positions to maintain relative to a leader or neighbor.

• No obstacle avoidance and no navigation. Chen and Luh:

• The analysis of group dynamics and stability.

• No obstacle avoidance and no navigation.

Mataric:• Investigated Emergent group behavior.• Simple behaviors can create an emergent

flocking behavior in group• Related to Reynols’ work

Parker:• Coordination of multiple heterogeneous

robots.• “Bounding overwatch”

II. Approach a) Line

b) Column

c) Diamond

d) Wedge

Formation maintenance in two steps:• Detect-formation-position: Determines

robot proper position based on current data.

• Maintain-formation: Generates motor command to correct robot location.

AuRA case: Motor schema control, The command is movement vector.

UGV Demo II: Separate votes are cast for steering and speed correction

1) Unit-center- referenced

2) Leader-referenced

3) Neighbor-referenced

Formation position determined by various referencing techniques

Each robot determines the positions of its peer by direct perception of other robots. GPS, Dead-reckoning

The robot transmit their current position in world coordinates with updates as rapidly as required for the given formation speed and environmental condition.

Position errors and latency are significant factor.

Nomad 150 simulation• Motor schema implementation.

UGV implementation• Decoupling of motor control into separate

steering and speed behaviors.

III Motor schema-based formation control Motor schemas:

• Move to goal, avoid obstacle, avoid robot, noise, and maintain formation.

Determined robot behavior:• Multiplying the outputs of each primitive

behavior by its gain, then summing and normalizing result.

Maintain-formation schema generates a movement vector toward the desired position.

The vector magnitude depends on how far the robot is away.

Ballistic zone, controlled zone, and dead zone

0%10%20%30%40%50%60%70%80%90%

100%

The deadzone provides a stable target area. Mobile robots require a small deadzone to

avoid oscillations about the formation position due to latency in communication.

The formation behavior is one component of the robots’ overt actions.

The resultant action depends on the relative strength of the formation behavior to the other goal-oriented behaviors.

A. Motor schema results in simulation Georgia Tech’s mission lab robot

simulation. Five sided polygons: robot Black circles: obstacle Solid line: Robot path Maintain position process:

Obstacle information request of that data send to simulation process A list comprised of angle and range data returned A robot moves and maintains the formation

Four robots in leader-referenced

Comparison of leader-referenced and unit-center-referenced diamond formation

B. Motor schema performance in turns Evaluation in turns:

• The robots commanded to travel 250m, turn right, then travel another 250m

• 90 degree turn• No obstacles• 10 simulations• Robot travel 100m before the evaluation

starts• 500m course evaluation excluding the 100m

initial travel

Path length ratio: (average distance traveled by the four robots)/(the straight-line distance of the course)

Position error: Average displacement from the correct formation position thru the run

Time out of formation: The percentage time out of formation data

C. Motor schema performance in an obstacle field Performance also measured for four

robots navigating across a field. Obstacles are placed randomly so that

2% of the total area is covered with 10-15m diameter obstacles.

Data from runs on 10 random scenarios were averaged for each data point.

The best performance: column formation In most instances, unit-center-referenced formations fare

better than leader-referenced formation Overall path length for robots in a leader-referenced

formation is generally longer than in unit-centered formation

D. Motor schema results on mobile robot

Two-robot team of Nomad 150 robots used: Shannon and Sally.

The robot were directed to navigate from West to East across the room.

Separate runs were conducted for each type of formation with and without obstacle.

Traveling in column formation

Traveling in wedge formation

Traveling line formation

IV. Formation control for the UGV demo II architecture UGV Demo II is aimed at fielding a

robotic scout platoon for the Army. Each UGV is a High Mobility

Multipurpose Wheeled Vehicle (HMMWV).

The UGV differs from the motor schema method.

• The behaviors are coordinated by speed and turn arbiters.

• Separate motor behaviors are developed for the speed and turning component of the behavior.

Non-holonomic constraint robot’s heading during formation impacts its ability to remain in position.

Positioning:• Fore-aft correction:

adjust speed• Side-side correction:

adjusting heading

Speed selection:1) If the robot is in formation, the best speed for

maintaining that formation is the current speed.2) If the vehicle is behind its position, it should speed up3) If the vehicle is in front of its position, it should slow

down.4) The selected change in speed should depend on how

far out of position the robot is.5) Since the speed arbiter implemented in the Demo II

architecture selects the lowest speed vote of all the active behaviors for output to the vehicle, formation speed control is only possible by slowing down.

For steering:1) If the robot is in formation, the best heading for position

maintenance is the formation axis.2) If the robot is out of position laterally and the formation

is moving, it should turn toward the formation axis with an angle that depends on how far out of position it is.

3) If the robot is out of position and the formation has stopped moving, the robot should head directly toward its postion.

Rpos,Rdir=robot’s present position and heading; Rmag=robot’s present speed; Fpos=robot’s proper position in formation; Faxis=formation’s axis Hdesired=desired heading dheading=computed heading correction dspeed=computed speed correction Vsteer=steer vote, representing the directional

output of the motor behavior. Vrmsspeed=speed vote, the speed output of the

motor behavior.

A. UGV behaviors for formation Demo II architecture requires a decomposition

of control into separate steering and speed control components.

The maintain-formation-speed:Vspeed = Rmag + K*dspeed (Rmag = robot present speed)

• Ballistic zone: dspeed=1.0• Controlled zone: dspeed=1.0~0.0• Dead zone: dspeed=0.0

The maintain-formation-steer:Hdesired = Fdir – dheading

Hdesired = Fpos – Rpos

Vsteer = Hdesired – Rdir

The magnitude of dheading determined as:

• Ballistic zone: 90 degree• Controlled zone: 0~90• Dead zone: 0

speed correction heading correction

V. Results for UGV demo II mobile robots

The behavior were implemented using Single-robot simulator.

The behavior were debugged by generating Artificial fixed trajectory for one vehicle, then observing simulated robot’s attempt to maintain position.

VI. Conclusion Reactive behaviors for four formation and

three formation reference type were presented

AuRA implementation is conceptually simpler and applicable to holonomic robots.

UGV implementation addresses the additional complexity of nonholonomic vehicle control.

In most cases, unit-center ref formations perform better than leader-ref.

Thank you!!!Have a nice weekend!!!