CollectiveForagingExperiment

Prev: Reaching and Grasping on a iCub humanoid robot	Up: Home	Next: Sensory-Motor Coordination

Table of contents

• Collective Behavior and Swarm Robotics

Collective Behavior and Swarm Robotics

FARSA allows to carry on both individual and collective experiments, i.e. experiments in which robots are situated in an environment containing other robots. The term "swarm" is generally used to indicate situation in which the number of robots concurrently situated in the same environment is large.

Setting up collective experiments is extremely easy. For experiments based on the EvoRobotExperiment class, it only requires to set the Experiment->nagents parameter to the desired number.

The CollectiveForagingExperiment plugin, for example, enables you to carry on experiments involving a colony of 10 MarXbots that can be evolved for the ability collect “food” and to bringing it back to the “nest” (which are indicated by the two blue cylinders in the Figure). The robots might cooperate to achieve better performance with respect to robots that operate individually. In particular the robots might coordinate to collectively explore the environment and to overcome the limitation of their individual sensory systems (i.e. that enable them to perceive the food area and the nest area only up to a limited distance). Indeed, the robots evolved for the ability to reach the food area and to bring the collected food to the nest area tend to form a dynamic chain between the two cylinders that enables them both to preserve information concerning the relative position of the two target destinations and to travel directly back and forth between distant locations that cannot be perceived directly (see Sperati, Trianni & Nolfi, 2008).

The plugin also allows to carry on more complex experiments in which the foraging robots also need to coordinate to help halted robots and to escape predators (indicated in the Figure in red). For more details see Ferrauto, Simione and Nolfi (in preparation).

In this experiment the colonies of robots are formed by fully-related genetic individuals (i.e. by 10 clones of the same individual genotype that give rise to 10 identical robots with 10 identical neural controllers).

References

Ferrauto T., Simione L., Nolfi S. (in preparation). Evolution of Cooperative Behaviours in Robotic Swarms: Environmental and behavioural richness create the condition for prolonged innovation phases.

Sperati V., Trianni V., Nolfi S. (2011) Self-Organised Path Formation in a Swarm of Robots. Swarm Intelligence, 5:97-119.