A single neural network (SNN) is often used as a control system for an autonomous robot. We propose Cooperative neural network Ensemble to design the Control System (CECS) of an autonomous robot due to the inability of SNN on the way of capturing the real world environment. We describe simulations on populations of neural network (NN) ensembles in two different situations where single neural network (SNN) controller and simple neural network ensemble (SNNE) are not effective. Firstly, we report simulations using a set of SNN controller. Secondly, we report results that the proposed architecture produces functionally different groups of weights for different individual neural network (INN) in the controller by means of a correlation function. We minimize a correlation function during the course of evolution which produce functionally different INN and weight vector for each INN. The robot can clearly differentiate its movements, following its either left side of right side environment all the way around. We confirm the results by a temporal correlation map (TCM) of the outputs of multiple neural networks.
[1]
Stefano Nolfi,et al.
Learning and Evolution
,
1999,
Auton. Robots.
[2]
Karl Sims,et al.
Evolving 3d morphology and behavior by competition
,
1994
.
[3]
Xin Yao,et al.
Ensemble learning via negative correlation
,
1999,
Neural Networks.
[4]
Randall D. Beer,et al.
Sequential Behavior and Learning in Evolved Dynamical Neural Networks
,
1994,
Adapt. Behav..
[5]
Francesco Mondada,et al.
Mobile Robot Miniaturisation: A Tool for Investigation in Control Algorithms
,
1993,
ISER.
[6]
Randall D. Beer,et al.
Evolving Dynamical Neural Networks for Adaptive Behavior
,
1992,
Adapt. Behav..
[7]
Francesco Mondada,et al.
Evolutionary neurocontrollers for autonomous mobile robots
,
1998,
Neural Networks.
[8]
Francesco Mondada,et al.
Evolution of homing navigation in a real mobile robot
,
1996,
IEEE Trans. Syst. Man Cybern. Part B.