SensoryMotorCoordination

Prev: Collective Behaviour and Swarm Robotics	Up: Home	Next: Body Evolution and Morphological Computing

Table of contents

• Sensory-Motor Coordination

Sensory-Motor Coordination

Sensory-motor coordination refers to the ability to act so to later perceive useful information. By coordinating their perceptual and action capabilities, robots can access and generate the information they need to carry on their task and can solve problems through solutions that are significantly more parsimonious with respect to solutions that do not exploit sensory-motor coordination.

The KepheraDiscriminationExperiment plugin described in this page represents a demonstration of how a problem that apparently requires to discriminate between walls and cylinders can be solved through a simple solution that does not require to categorize the two type of objects. The possibility to identify this simple solution is crucial to solve the problem, since the stimuli perceived near the two types of objects largely overlap in the robot’s sensory space (Nolfi, 2002, 2005).

The AbstractDiscriminationExperiment plugin allows to replicate a simplified version of the experiment reviewed above in which an agent that is situated in a circular environment and that can move only clockwise or counterclockwise needs to reach and remain in the left side of the environment. The environment is constructed so that each of the 20 different stimuli that the robot can perceive are present both on the left and on the right portion of the environment. Consequently, the perception of any stimulus by itself does not provide any information on whether the robot is located on the left or right side. Despite of that, the agent can solve the task on the basis of a reactive controller that does not have the possibility to remember previously experienced stimuli (Nolfi, 2002, 2005).

Finally, the KheperaNavigationExperiment plugin allows to replicate an experiment that shows how a robot can coordinate its sensory-motor activity so to generate and use information that it cannot perceive from the environment. In this case, a Khepera robot placed in a rectangular environments in which the length of the top and bottom walls exceed the length of the left and right wall is asked to navigate and remain in the top-left or bottom-right corners of the environment while avoiding the other two corners. The discrimination of the target corners with respect to the other two corners can be carried out by discriminating the length of the walls which however cannot be perceived by the robot on the basis of the available infrared sensors. The experiment shows how, by coordinating the sensory and motor process, the robot can solve the problem through a simple strategy that consists in reaching a corner and then leaving it with an angle of about 45 degrees from the two walls forming the corner. This in fact ensures that by turning left and then following the wall, when the robot encounter a wall on its left side, it will always navigate toward the two right corners. In other words the strategy of abandoning a corner with a 45 degrees angle enables the robot to infers whether the walls encountered later on are long or short (walls encountered on the left side are long while walls encountered on the right side are short). Indeed, turning left and then moving forward along walls encountered on the left side, enables the robot to always travel toward the top-left or right-bottom corners (Nolfi, 2002, 2005).

References

Nolfi S. (2002). Power and limits of reactive agents. Neurocomputing, 49:119-145.

Nolfi S. (2005). Categories formation in self-organizing embodied agents. In H. Cohen & C. Lefebvre (Eds), Handbook of Categorization in Cognitive Science, pp. 869-889