Truck backing up neural network controller optimized by genetic algorithms
M.L. Ho, P.T. Chan, Ahmad B. Rad, Chung Hong Mak · 2004
Jenkins-Yuhas network has been applied successfully to solve the trailer-truck backing up problem. Genetic algorithms (GA) is used to train the Jenkins-Yuhas network controller. The integration of genetic algorithms and neural networks training avoids complicated formulation of derivative function that required in conventional error back propagation techniques (gradient descent). In addition, it can avoid trapping in local minimum point. The performance of GA trained neural network controller is evaluated via simulation studies. It has been demonstrated that the controller can successfully control trailer-truck for different initial parking conditions, i.e. with same set of trained weights, to the loading dock.