vehicle in our project is a 1999 Toyota
Landcruier 4WD. A 4WD
vehicle was chosen for a number of
reasons: it provides a strong and robust platform capable of surviving
the rigors of experimentation; it has a large amount of interior space
for installing sensors/computers; and it allows the option
of performing research into off-road autonomous
driving. The overall design philosophy is to use as many off the shelf
components as possible to reduce development time and to lever off
existing tried and true technology.
The main mode of
sensing used in the vehicle will be vision.
Two separate vision systems are planned. First, an active
vision head (called CeDAR developed previously at the ANU - see active vision page)
be mounted with two stereo camera pairs. One pair will have a short
focal length, and concentrate on the near field of view, while the
other pair will have a longer focal length, and concentrate on looking
further along the road. The second vision system involves using a
pair looking from the dash back toward the driver's face. By
monitoring the driver useful information as to their intention can
be gathered as well as verification that they have seen a detected
dangerous situation. This system is based on the faceLAB system from seeing machines. Apart
from vision sensing, a Global Positioning System (GPS), Inertial
Navigation Sensor (INS), and laser range finder have been installed
vehicle. The 6 DOF INS is mounted close to the vehicle's centre of
gravity at a point between the two rear-seat foot-wells. It provides a
continuous stream of linear and angular acceleration data that can be
used to keep track of vehicle dynamics. The GPS provides data that
can be used for high-level, navigation problems, but is also very
useful for correcting drift in the INS output. The laser range
finder has been mounted looking forward on the vehicle's bull-bar. Its
purpose will be to identify obstacles, both stationary
(eg. guard-rails, parked cars, etc.) and moving (eg. other vehicles),
and will provide an additional source of information for our obstacle
sub-systems are required in the vehicle:
steering, braking, and throttle. We achieve throttle control by
interfacing with the vehicle's cruise control module. The steering
sub-system is based around a Raytheon rotary drive motor/clutch unit,
which was designed for use in yacht auto-pilot applications. It was
the engine bay alongside the steering shaft of the vehicle. Power from
an electric motor is transferred to the steering shaft using three spur
the first is attached to the steering shaft, the second to the motor
shaft, and the third, being an idler gear, sits between the first
two. A key feature in the design is that the idler gear can be engaged
and disengaged from the drive-train using a lever protruding from the
assembly. Then for ``manual'' driving of the vehicle, the idler
gear can be disengaged, providing the safeguard that the autonomous
steering assembly cannot impede normal steering in any way. A
photo of the steering sub-system is shown in this image.
Note the lever used to engage and disengage the idler gear. Also note
the rotary drive motor/clutch unit, and the vehicle's steering shaft.
The braking sub-system is based around a linear drive unit (produced
by Animatics), and an
electromagnet. The linear drive is connected to one end of a braided
steel cable via the
electromagnet. The cable passes through a guiding sheath to reach, at
its other end, the brake pedal. Braking is then achieved by having the
linear drive unit pull on the cable. The electromagnet must be powered
in order for braking to occur (ie. if it is
unpowered, then the linear drive cannot pull on the cable to activate
the brake). In an emergency, power can be cut to the
electromagnet so that all braking control is returned back to the
driver. In our implementation, an emergency scenario is communicated
to the autonomous driving system by having the human activate an
emergency stop button. The braking subsystem is shown in this image. In the foreground
the figure shows the linear drive
and electromagnet, while in the background the brake pedal and its
connection with the cable is shown.
communication hardware is required to fuse
together the various sensing and actuation subsystems into a cohesive,
single unit. Our approach in this area has been to favor the use of
standard PC and networking hardware. Such hardware is readily
available, easily upgradable, and cheap. An additional
PC will be installed to process non-vision sensing data, and to
control the throttle, steering, and braking subsystems. Communication
is achieved between PCs via ethernet, with a connection from the
vehicle back to a base station possible via a radio ethernet link. Due
large number of sensing and actuation devices that communicate over
serial lines, a serial port server has been installed. This device
allows communication between a PC and serial devices as though these
devices were connected directly to local serial ports on a PC.
Finally, an Servo to go card
installed to provide a low level communication interface between PCs
and various other devices (eg. cruise control system, steering motor
control, steering angle potentiometer). This module connects into the
ethernet, and provides a
number of functionalities, including A/D and D/A conversion, PID
control, timers, etc.
Cruiser, diesel, with power steering,
cruise control and ABS.
brake actuator system.
Smart motion linear actuator.
Steering mechanism of vehicle including drive motor/clutch unit (left),
idler gear (centre) and gear on steering shaft (right).
|CeDAR active camera
FaceLAB cameras on dash board.
LCD monitor in backseat with FaceLAB software on screen.
|| Sick Laser Range
Mbit Radio ethernet traceiver.
| Looks like a 10cm3
rate and acceleration gyroscope.
| Size of a bread box
DC - 240V AC inverter.
| Cylinder with
cable coming out one end.
angle linear potentiometer.
|looks like every
other ethernet switch.
The software runs on several machines running linux.
based software architecture written in C/C++ and CORBA.