A general steering strategy was developed to navigate
the tractor on sloping terrain, where a neural network
model was trained for four slopes of different gradients,
respectively. The corresponding steering controllers
were then developed but were generalised for the sloping
grounds of any inclination. From an autonomous travel
test it was found that for the tractor navigation by the
developed guidance system, the mean and standard
deviation in tracking a contour path on 208 sloping
ground were 005 and 00439 m, respectively, whereas
that for the navigation by separate reference table
method were 01464 and 00667 m, respectively. It was
also found that the mean and standard deviation in
tracking the contour path of 118 sloping ground by the
separate reference table were 00267 and 00338 m,
respectively, whereas those by the general reference
table method were 0028 and 00297 m, respectively.
From the results it is proved that the developed
guidance system can precisely navigate the prototype
tractor along rectilinear contour paths on sloped
terrains of any inclination, whereas the separate
reference table method gives better results on specific
trained slope. The general reference table method could
save time to attain a total navigation on sloping
environment.