Abstract
There is found a semiactive suspension neurocontroller based on elementary perceptron model and using only input-output information about the system. Consequently, the structure of neurocontroller is implementable, supposing an online training of a physical system. A comparison is made both with the active LQG and the conventional, sequential derived, systems. The obtained performances involving well known suspension system indices (safety and comfort indices) are promising. The novelty of stochastic framework of the learning semiactive algorithm marks the principal success of the paper.