Cooperation in swarms of robots without communication
Jianing Chen
Cooperative transport based on occlusion
View the summary of work and experiment videos.
Segregation based on the Brazil nut effect
View the summary of work and experiment videos.
Fully decentralised segregation
In Chapter 5, a decentralised controller was implemented to organise a swarm of e-pucks that allows e-pucks to mimic the behaviour of particles.
We showed that a swarm of robots representing particles of two to three different sizes can segregate into an annular pattern.
The motion of a robot was governed by a combination of three components:
- Attraction towards a global location, which emulates the effect of a gravitational pull.
- Random motion, which emulates the effect of vibration.
- Repulsion from nearby robots, which emulates the effect of collisions between the particles.
Although this controller only used information from a robot's own sensors, the presence of a global reference point and the assumption of all robots to perceive it may not be met in practical applications.
In this chapter, we introduce a new algorithm to form an annular structured segregation that does not require a global reference point.
In this algorithm the motion of a robot is still based on the three components in the Brazil nut effect. However, the attraction towards a global location was achieved through a collective signalling mechanism that does not involve any global information.
Focusing on this new attraction mechanism, simulation was implemented to study the performance of this controller.
Technological platform
The technical development for the e-puck robot in this thesis has been organised into a library, including the black shirt and the colour cap.
View the distribution page of this library on GitHub.