Loading...

Control and Optimization of Robotic Swarm Flocking Using Naturally Inspired Methods

Vatankhah, Ramin | 2009

931 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 39808 (08)
  4. University: Sharif University of Technology
  5. Department: Mechanical Engineering
  6. Advisor(s): Alasty, Aria
  7. Abstract:
  8. The idea of using groups of simple autonomous robots instead of one or limited number of very sophisticated robots is inspired by flocking behaviors of animals. Distributed control structures and artificial swarming have attracted lots of studies in robotics. There are many applications for robotic swarms in which members are very simple robots with low communication capabilities. These robots communicate only within very limited distances. In this thesis, a homogenous robotic swarm with at least two robots is assumed. Swarm robots are considered to be dimensionless with no time delay, which is a normal assumption in this field of studies. Robots move in planar space and their behaviors are result of two different phenomena: interactive mutual effects (combination of both attraction and repulsion) between robots and influence of the Leader-agent. To be closer to real conditions, the fields of the robots' view are limited. The Leader-agent is the only one who has extra communication abilities and can determine its position relative to destination. So every robot’s motion also may be influenced by the Leader-agent. A robot feels attraction toward the agent if they are not farther than a specified distance. Influence of the Leader-agent on the robotic swarm would be local. In this thesis, we consider a control strategy of the robotic swarm based on reinforcement controllers named adaptive critic-based neurofuzzy control technique. Robots as autonomous members may have higher equations of motion which can violate coordination control algorithm of the robotic swarm. To force every robot to behave as a member with quasi-static equation of motion we have designed a behavioral controller. The controller is implemented and results are verified via simulation. To implement the Lyapunov control algorithm, maximum possible velocity of the swarm should be determined. Determination of this value is not mathematically possible. An online intelligent optimization subsystem is needed. Using naturally inspired methods, the Leader-agent velocity is determined to maximize the swarm velocity. Particle swarm optimization, ant colony optimization and genetic algorithm, are used to achieve this goal. Results show the high ability of these evolutionary algorithms in solving complicated dynamic optimization problems.

  9. Keywords:
  10. Genetic Algorithm ; Particles Swarm Optimization (PSO) ; Ant Colony Optimization (ACO) ; Robotic Swarm ; Naturally Inspired Method ; Reinforcement Controller

 Digital Object List

 Bookmark

No TOC