This paper describes lateral and longitudinal control algorithms
when vision-based autonomous vehicles form a platoon by detecting a
preceding one with a machine vision system. The visual platooning
is an extension of the longitudinal control in ACC, and is featured
by both the lateral and longitudinal control of following vehicles.
The lateral control algorithm for a following vehicle uses a target
point positioned on a preceding vehicle and captured in the field
of view. When the target point is appropriately positioned, a
following vehicle is guided as if it chased the target.
The origin of the lateral control algorithm is an algorithm for
an automated vehicle with a localization function including the
dead reckoning, named the target point following algorithm. The
lateral control in the visual platooning consists of the steps of
target point detection, path generation, and lateral control
calculation. A coefficient of a cubic curve that is defined with a
target point and the current position of the following vehicle
provides the steering angle. The path for the following vehicle is
not necessarily the same as the preceding one, because the target
can be positioned with freedom to some extent.
The longitudinal control algorithm of a vehicle is determined
with a distance between a preceding vehicle and a following one.
The vision system or the laser radar on the following vehicle
measures the distance to the preceding vehicle. The speed control
of the following vehicle is calculated with the distances measured
during the previous control cycle and the present cycle, the
present speed of the vehicle, the reference distances and the
duration of a control cycle.
The feasibility of the algorithms is shown by simulation studies
and experiments with indoor mobile robots and automated passenger
cars on a test track.