Hierarchical and Modular Fuzzy Architecture for Multiple Mobile Robots
Implementation of autonomous behaviours in mobile robots, using fuzzy logic control, requires formulation of rules that are collectively responsible for necessary levels of intelligence. This collection of rules can be conveniently decomposed and effectively implemented as a hierarchy of fuzzy behaviours. For this purpose, a hierarchical approach to the incremental design of complex robot behaviours based on fuzzy logic is proposed. Since the primitive behaviours of individual robots have different aims and this may cause conflict. Some of these behaviours are competitive and some of them are cooperative. This paper also proposes a new hybrid behaviour coordination system between the competitive and cooperative methodologies with the aim of taking the advantages of robustness and optimized robot trajectories from both approaches. Simulated and real experiments have been conducted on team of mobile robots performing a proof-of-concept dynamic target tracking task. The results show an improvement in overall group performance even in cluttered environment.
| Attachment | Size |
|---|---|
| PAPER105AUDIO.wmv | 4.17 MB |
Dear sir
Thank you for your very good questions
First, the target is movable not fixed
Sencod, the aim of the robots is to cooperate to track and capture the target, so if this happened, the job is finished which is a kind of security application.
Third the system is reactive because it works in dynamic environment and the target motion trajectory is not defined apriori.
Finally the subsumption architecture is competitive architecture based on the priorities of the behaviours but there are some situations which require the robots to perform more than one behaviour at the same time such as avoid obstacles while following a target. In subsumption architecture this can not occure. However, fuzzy logic which is a cooperative approach can mix both behaviours and the output will be the resultant of both behaviours based on the wieght of each behaviour.
Thank you for your contribution to IPROMS2007.
I would ask some simple questions about the initial status of the practical experiments. What is the dimension of the area for the mobile robot?
What is the maximum distance that the obstacle sensors can detect? At the initial status that the robots were placed at random places, will the time be different if they are placed farther or closer from each other, especially when they are placed at "blind" locations?
Many thanks.
Regards.
Dear Ji
For the initial conditions
The area was about 2 m^2
The obstacle sensors range was about 30cm
Of course yes if the robots are far from the target, it will take longer time to track the target, the experiments were run thirty times and the results are averaged, aslo some few trails the robots failed to track the target especially if the environment was full of obtacles.
Would it be possible to upgrade this system to follow a mobile target?
The three robots capture (or reach) the target, what do they do with it afterwards? Doing something to it afterwards would require real collaboration.
A reactive agent is meant to overcome the issues of symbolic AI, some might consider Fuzzy Logic to be a part of symbolic AI. Does that make your system truely reactive?
What benefits does a fuzzy reactive architecture offer over other approaches such as the subsumption architecture of Brooks?