-----------------------------------------------------------------------------------------------
This is a summary of the Article “Distributed Online Evolution for Swarm Robotics”
Hettiarachchi, Suranga. Distributed Online Evolution for Swarm Robotics Print.
-----------------------------------------------------------------------------------------------
This article discusses some difficulties and challenges in creating a swarm robot system such as noise, limited range of interaction between agents, delayed feed back and many others. Also, solutions to such problems were discuss such as using evolutionary algorithms (EAs). The fact the these algorithms are created and worked in simulation programs does not prepare the system for real life problems, also creating different individual behavior for different agents becomes a problem.
DAEDALUS “Distributed Agent Evolution with Dynamic Adaptation to Local Unexpected Scenarios”, a new interesting framework was presented in this article. This particular paradigm depends mainly on the interaction between the agents in the swarm system or outside it. Many different runs and experiments showed improvement in the swarm system in both collision avoidance and and survival separately. Future research is interested in combining both aspects together “ Collision avoidance can be improved via mutation of the obstacle-robot interaction, while survival can be improved via mutation of the robot-robot interaction and robot-goal interaction”. Also Future research is focusing on creating a credit system in order to reward robots based on their own achievement.
Time: 4hrs.
No comments:
Post a Comment