Central action selection using sensor fusion

A computational model of central action selection has been successfully embedded in a robotic platform. The proposed model, allows action selection between basic behaviors in a rat-like fashion. These behaviors imitate some home cage activities of a laboratory rat. The behavioral setting ranges from: search-´food´, ´food´-pickup, ´food´-deposit, wall-finding and look-around. There is evidence that the switching of multiple sensorimotor systems in the vertebrate brain is similar to the concept of action oriented sensor fusion in AI robotics. In this work, by integrating infrared, odometry and visual information within a reactive navigation scheme, we use sensor fusion to produce action selection. Following this approach, initial experiments were conducted using a robot simulation software. Subsequent tests were carried out to evaluate the functionality of the model in a Khepera robot.