Hybrid self-organizing map and locally recurrent neural network-based adaptive back through for improving integrated vehicle stability control

If the variable of stability such as Yaw-Rate (YR), Vehicle Side Slip (VSS), Roll Angle (RA) cannot fulfil desired value, then occur instability vehicle direction causing accident. Four preceding stability integration controls, feed-forward control, H^∞ control, nonlinear predictive control and robust control, could not perform desired target, yet they were not able to adapt the driving conditions covering vehicle dynamic, drivers character and environment called three in one dynamic system (TODS). Hybrid multi dimension fuzzy C-mean cluster-adaptive back through control (MDFC-ABC) could either control TODS adaptively or reduce more error than those of the four, but it has exploited both architecture and computational work. In this paper is proposed new integration control design based on combined kohonen feature map-associated memory and locally recurrent neural network-based adaptive back through control (KFMAM-LRNNABC). LRNNABC consists of NNPlant and NN-Controller. Architecture of NNs are resulted from locally recurrent neural network (LRNN) with a hidden layer. The methodologies, first step it generates TODS pattern data realized by the test drive of ESP-4WS-AS vehicle, that are normalized. Each of 17.280 normalized data entries to KFMAM input-output layer to map the winner and its neighbourhood of 540 neurons in MAP layer iterated to become fixed weight matrix. Second step, the training process to update the optimized architecture and parameters of NN-Plant/control based on constructive back propagation (CBP) uses the fixed weight matrix of KFMAM. An experiment and simulation was completely ran to compare the performance of hybrid KFMAM-LRNNABC integrated control and hybrid MDFC-ABC in controlling TODS. The simulation result shows that performance hybrid KFMAM-LRNNABC was better than those of hybrid MDFC-ABC.