Anders Rehan Pattern

Non Critical Bugs

• Changes direction of go around (phantom obstacle?)
• Search for connect leds random walk needs to be more noisy

Experiment Log

Star

Overhead star_20070226_1

• experiment1 - 8 did not have correct optical barrier threshold set
• experiment 2 - sbot fell over and was righted by anders

overhead camera pattern_line_20070227_1

• 3 runs with line, handshaking still used.

overhead camera pattern_line_20070227_1

• Experiment 5. Anders rights toppled sbot 32.
• Experiment 10. Sbot 6 Didn't start. Restarted.

square_20070228_1

• Exp 1 repeated - sbot 23 crashed.
• Exp 2 repeated - sbot 23 crashes during handshake. Logs saved.
• Exp3 repeated - sbot 8 fails to start
• Exp 4 Ooorf connection.

Experiment 2b redone

• Arrow exp 5 - sbot 8 had incorrect optical barrier set and was restarted

==E

A team

• 13
• 32
  • Focus?
  • Reoptimise Cols?
• 24
• 30
• 6
• 36

Seeds

• 35
• 22

Rejected from A team

• 22
  • One of two gripper leds not working

Sbot Vals

Optical Barrier Value

Sbot 8 - 30 Sbot 32 - 25 Sbot 23 - 60

Particularities

sbot 35 - Optical Barrier gives too high values when over leds. Can't use in experimetns except as seed.

Paper Structure

All papers have common thread of functional pattern formation.

Paper 1

Pattern Formation in a Swarm of Self-Assembling Robots

• What patterns can we make.

Simulation

• Analysis of patterns.
• Simple (very abstracted) demo of functional pattern formation.
  • Trigger could be environmental (green floor) or external command (sound).

Real Robot

• Proof of concept. Working demo. Doesn't have to be reliably working.

Paper 2

Function Dependent Pattern Formation in a Swarm of Self-Assembling Robots

• What can we do with our patterns
• Which patterns are better suited to which tasks

Real robots

• Pre assembled patterns tested against various tasks / obstacles
  • Hill
  • Trough
  • Moving S-Toy
  • Rough Terrain

Simulation

• Functional self-assembly working in a realistic (non-abstracted way).
• Robots detect obstacles / tasks and perform appropriate patterns as determined above.

Paper 3

Deployment of Functional Dependent Pattern Formation in a Swarm of Self-Assembling Robots

• More realistic task - Lots of robots / Ongoing

Simulation

• Completely distributed algorithm.
• Probabalistic seeding of multiple, independent patterns

Real

• Functional Pattern formation in a real task. e.g. Trough - Hill - SToy.

Reference Papers

Pattern formation in mobile actuator and sensor network Chen

Example of real world robots using (pseudo) gps to form patterns using simple robots. Each robot is related positionally to leader who broadcasts his position.

Pattern formation and optimization in army ant raids Bonabeau

Example of functional pattern formation in the natural kindom. Mathematical modelling.

A general algorithm for robot formations using local sensing and minimal communication. Fredslund, Mataric

Simulation and Real robots. Formations formed by keeping angle and distance from friend. Global communication of desired pattern paramters. Only local sensing. No global coordinates. Robots have pre-assigned id's and positions in formation.

Social patterns for scalable multi robot formations. Balch Hybinette

Simulation study. Fixed formation. Not fixed positions within formation. Motor schema vector summation. Performance measures of different formations by crossing obstacle filled arena.

Cellular Robotic Systems - Self Organizing Robots and Kinetic Pattern Generation Wang Beni

Grid space. Rules to generate patterns. Generic algorithm to generate given pattern. Guaranteed to terminate.

Superlinear physical performances in a SWARM-BOT Mondada

Discusses optimal structures (size rather than shape) for various tasks and the relative performance increase of adding subsequent sbots. Could be used to justify i) Star shape chain size ii) Group size selection mechanism.

SWARM-BOT - Pattern formation in a swarm of self-assembling mobile robots. Sahin

Swarmbot project. Simulation only. Hexagonal grid world. Probabilistic assembly. Statistical analysis of structures formed. e.g. length of chains, frequency.

Steps towards self-reconfigurable robot systems by modelling cellular adhesion. Ottery, Hallam

Simulation only. Simple heirarchical patterns based on cell membrane type interactions (A-CAM). Paper we reviewed was partially based on this.

To Download

• Gradual spatial pattern formation of homogeneous robot group. Fukuda. (Science Direct)