We showed in a previous work that an artificial evolutionary system whose task was to track a light source was able not only to evolve and grow a neural network, but was also able to evolve learning mechanisms. The evolved neural network was then transferred to a robotic system consistent of a camera mounted on the gripper of a robot arm with results comparable to the ones achieved by the simulation (Eggenberger et al.[1]). In this paper we continued testing the evolved controller in the real-world increasing the sensorymotor capabilities and the robot’s task as follows: (a) The visual system was enhanced to detect color and movement in the environment and a proprioceptive system was added to have feedback of the arm movements. (b) The number of degrees of freedom (DOF) of the robot was increased from two to three. (c) The position of the cameras was fixed and the same underlying principles were used to teach the robot arm in front of the cameras to move a colored object from an initial location at the periphery of the visual field to the center of it. The arm could solve the task not only for two DOF, but also for three DOF.
[1]
Olaf Sporns,et al.
Neuromodulation and plasticity in an autonomous robot
,
2002,
Neural Networks.
[2]
P. Eggenberger.
Evolving neural network structures using axonal growth mechanisms
,
2000,
Proceedings of the IEEE-INNS-ENNS International Joint Conference on Neural Networks. IJCNN 2000. Neural Computing: New Challenges and Perspectives for the New Millennium.
[3]
P. E. Hotz,et al.
Evolving the morphology of a neural network for controlling a foveating retina: and its test on a real robot
,
2002
.
[4]
Karl J. Friston,et al.
Value-dependent selection in the brain: Simulation in a synthetic neural model
,
1994,
Neuroscience.
[5]
Akio Ishiguro,et al.
Neuromodulated Control of Bipedal Locomotion Using a Polymorphic CPG Circuit
,
2003,
Adapt. Behav..