Learning of biologically inspired behaviors for autonomous robots by a navigational network
Paulo A. Jiménez, Bijan Shirinzadeh, Yongmin Zhong · 2008
Many animals can navigate by means of Path Integration (PI), in which an animal keeps a continuously updated record of its current direction and distance from some reference point as it moves away from that place. Equipped with such knowledge, a navigational network learned how to maintain a vector pointing to the home base, with a precision that depends on the number of neurons used to encode the vector. As PI systems are prone to cumulative navigational errors, the robot must perform a systematic search for the home base once it has reset its path integration system and yet not arrived at the home base.