Application of Neural Networks on Off-Road Vehicles

Off-road vehicle traction performance has been considered as an
important point of research for many years theoretically and experimeiitally,
where the design of traction device system requires developments of
mathematical models. One of these developments is directed to predict the
performance of vehicle traction. Recently, Artificial Netiral Networks (ANN's)
seems to provide an alternative approach to the modeling of off-road vehicle
performance. In this paper, an application used for computing the drawbar-pull
for off-road vehicle as a function of some input set of data vectors using a
suggested ANN's model is considered. The suggested model provided with
more simplified topology of ANN's architecture. The model comprises only a
single hidden layer of neurons with some dynamic controllability by
monotonous increasing of hidden layer neuron number. Comparisons between
the suggested model results and others proving the better performance of the
simplified model. However the relationships between pull and load at different
running conditions (inflation pressure, tyre size, and pull angle) have been
predicted using the ANN's model. Comparisons between previous model and
published data with ANN's predicted results have been made and the agreement
seems to be good. The model presented herein is a multi input-one hidden layerone
single output, and suggestion for fu􀎵lier development is, th􀎵
result model. Finally, a suggested direction of fhture, modification of our model
with more biologically plausible consideratiqn is given. The expected advantage
for such modified models is that they are more practically applicable to solve
complex engineering problems in the real world