Neural Network based Controller for High Speed Vehicle following Predetermined Path

The actual integration of automated control systems in vehicles such as Anti-lock
Braking Systems (ABS) or Traction Control System (TCS) has proved to increase road
safety and improve driver's comfort. Since most of the accidents are attributed to the
fault of the driver, automated control systems in vehicle safety technology may
dramatically better road safety by improving driver's performance. This thesis presents
an enhanced and improved autonomous intelligent cruise control systems with obstacle
collision avoidance integrated with path following/lane keeping. Obstacle collision
avoidance is the ability to avoid obstaclesthat are in the vehicle's path, without causing
damage to the obstacle or vehicle. Path following/lane keeping is the ability to follow
the vehicle's path and keeping in its lane, as accurately as possible. The idea is to have a
vehicle that drives by itself and avoids obstacles in the real world. Every instant, the
vehicle decides by itself how to modify its direction according to its environment. This
thesis demonstrates Gaussian functions and multi-objective cost function employed
alongside with the Neural Network and optimal preview controller for control of the
position of the vehicle to move while avoiding collision with obstacles. Each obstacle is
represented independent of the others as a bell-shaped hump by the Gaussian functions
which serve as an obstacle recognition system. Multi-objective cost function is formed
for the planning strategy to generate, evaluate and select plans so that the vehicle can
select which direction to move. Neural Network and optimal preview steering control
are utilized to control a full linear steering model of a vehicle so as to increase path
following accuracy. Optimal preview control is capable to portray the driver's vision of
the path and process the knowledge while Neural Network controller has the ability to
'learn' from past errors and adjust the network to obtain specific target output. In this
thesis, a MATLAB simulation environment was created to simulate the ability of a
vehicle to avoid obstacles that are in the vehicle's path. Simulated obstacle avoidance
has confirmed the capability of a vehicle to precisely avoid collision with obstacles
while traveling on high speed along its predetermined path.
in