This article was first published in 2007.
|
Cruise control systems have been available in cars for many years. However, a
new type of cruise control is now being fitted. It’s called Adaptive Cruise
Control and it uses radar to maintain a safe distance to the car in front, even
if that car’s speed changes. On the road it’s a brilliant innovation that
improves safety, reduces fatigue and adds convenience. But how does it
work?
Intelligent Cars
The last decade has seen the widespread introduction of systems than enhance
car intelligence. Anti-Lock Braking (ABS) and Electronic Stability Control (ESP)
give the car the ability to act in ways not specifically requested by the driver
– for example, to release the brakes momentarily to prevent wheel lock-up or to
reduce throttle opening if the car is sliding. Adaptive Cruise Control is
another step on that road to enhanced intelligence.
The presence of systems like ABS and ESP means that many of the input signals
needed by Adaptive Cruise Control are already available. These include:
vehicle speed
vehicle lateral acceleration
driver accelerator input
driver steering input
driver brake input
However, not present is the most critical of inputs – a forward-looking
sensor.
Forward-Looking Sensor Characteristics
An Adaptive Cruise Control forward-looking sensor must meet certain strict
design requirements.
1. Range
In order that an appropriate following distance can be maintained, the sensor
must be capable of working over a specific range. For example, if the following
distance is defined in terms of time gap between the two vehicles, a 2 second
gap at 160 km/h will require the distance between the vehicles to be about 90
metres. However, in order that the sensor can maintain continuous control, the
actual required sensor range will be about 10 per cent greater than this. So if
the maximum speed required of the ACC is 160 km/h, a sensor range of about 100
metres is the minimum requirement
2. Closing Rate
The sensor must be able to rapidly detect that the car ahead is being closed
upon. If the sensor is slow to react, a greater range will be required of it
otherwise the following car will draw too close before throttle reduction or
braking occurs. The magnitude of permitted braking will also affect this
requirement: if the car is permitted to brake hard then the sensor can be slower
to react. Assuming a maximum automatic braking deceleration of 0.2g, a maximum
closing rate of 50-65 km/h and a minimum following distance of 20-30m, a sensor
range of 80-100 metres is again a minimum requirement.
3. Field of View
The field of view (FOV) of the sensor can be defined both in terms of azimuth
(left/right) and elevation (up/down) angles. The azimuth FOV is important if the
system is going to be effective at working on curves. As this diagram shows,
beam width has a major affect on the distance at which a cornering car can be
tracked. At a speed of 90 km/h the ACC following distance will be about 50
metres. Assuming a minium radius-of-curvature of 300 metres, a minimum sensor
FOV of 5 degrees is required. However, addition FOV is usually needed to take
into account mechanical or electrical misalignment of the antenna – a point that
we‘ll come back to.
In addition to these three requirements, the sensor must be able to withstand
a temperature range of -40 - +80 degrees C, be proof against water splashes and
pressurised steam, be immune to vehicle vibrations, resist stone impacts and be
as small as possible.
Two types of forward-looking sensor have been developed – lidar and radar.
However, the radar-based sensor is most widely used and it is this type of
sensor that will be covered here.
Radar Sensors
Two types of radar sensors are used – those with stationary antennas and
those that mechanically sweep back and forth.
US automotive components manufacturer Delphi has developed a scanning sensor
with a narrow 2-degree beamwidth. This beam is mechanically swept over a
15-degree detection region and has an elevation FOV of 4 degrees. As the antenna
is scanned, over 40 individual transmit/receive beams are executed with each
pass.
However, much more common is a sensor that has a fixed antenna. The Bosch
system (used by DaimlerChrsyler, BMW and Audi) uses this approach.
The Bosch system uses a Frequency Modulated Continuous Wave (FMCW) output.
Instead of timing the period between transmission of the signal and the echo, a
FMCW radar system compares the frequencies of the transmitted signal and its
echo. The output frequency is changed at a rate of 200MHz per millisecond and so
the time interval between the transmit and receive signals can be established by
determining their frequency difference.
However, because the distance between the transmitter and its target may be
changing, this differential frequency information contains not only the time
interval component but also the frequency shift (ie Doppler component)
indicative of the change in distance. This ambiguity can be resolved by the use
of multiple FMCW cycles using differing rates of frequency change.
By these techniques the distance to the target and whether the target is
drawing closer or moving further away can be established. However, more data is
also needed – is the target directly ahead or to one side of the forward aim? If
the target’s radar reflective characteristics are known, the amplitude of the
signal echo depends on the angle at which the signal is received by the radar.
However, when the reflective characteristics of the target are unknown, a
different approach needs to be taken. To determine the angle at which the radar
detects an object, three radar lobes are transmitted and analysed. The ratio of
signal amplitudes of the three different lobes provides this angular
information.
The Denso radar used in Toyota/Lexus models uses a more conventional type of
radar. Distance is detected by measuring the time between transmission and
reception, while relative speed is detected by the frequency shift (Doppler
Effect) of the reflected waves. The angular position is detected by the phase
differences of the signals received by multiple antennas. The Denso unit also
differs from the Bosch design in that it has separate receiving and transmitting
antennas (although all the antennas are mounted in the one assembly).
The physical layout of the Bosch radar sensor is shown here. The radar and
the ACC controller are integrated into one housing. The front of the unit
features a Fresnel lens that is used to focus the three radar lobes. The lens is
made from a special temperature- and stone-resistant plastic which is formed as
part of the module casing. The lens incorporates a heating element which
prevents it becoming coated in snow or ice. According to Bosch, wet snow has a
particularly great attenuating effect on the radar signal.
In one iteration of the Bosch design the sensor assembly comprises three
circuit boards. The first consists of the radar transceiver unit which is
mounted directly on a circuit board, keeping interconnections as short as
possible and so reducing susceptibility to interference. Also on this board is a
digital signal processor, purpose-developed 10- and 12-bit analog to digital
converter, SRAM and flash memory. On the second board is a 16-bit
microcontroller which performs the necessary car speed control calculations. The
third board contains the driver modules to allow connection to the car’s
electrical and CAN bus communications systems.
The module must be aligned in both vertical and horizontal planes. In the
horizontal plane Bosch state that a degree of accuracy of better than 0.3
degrees is required, while BMW put the figure at 1 degree and Cadillac at 2
degrees. The BMW system requires the use of a BMW service tool to perform the
alignment, while Cadillac systems can be placed in an ‘alignment mode’ and then
automatically aligned by being driven along with a road that has stationary
objects either side. Apparently, the more stationary objects (such as light
poles, mail boxes, etc) there are, the quicker the self-alignment occurs!
How the System Works
It is all very well to detect the presence of cars in front, but how is it
determined whether the car is in your lane or another? What about when
cornering? And what happens when a car cuts into your lane?
This diagram shows the signal processing architecture of a typical ACC. Once
the objects are detected, tracking of them occurs. Both their paths and also the
path of the controlled vehicle are estimated, the input commands of the driver
are noted and the ACC controls the throttle and/or brakes.
In the Bosch FMCW system, positive detection of objects is carried out by
comparing consecutive radar modulation cycles. If in the second cycle the object
is found where on the basis of its previously detected speed and position it
could be expected to be placed, it is assumed to be the same vehicle. In other
words, the object data is filtered on the basis of historical information.
Additional object tracking functions are carried out where there are multiple
simultaneous echoes from different distances, which can be the case with large
trucks. In this situation the multiple echoes are combined so that the system
sees only one object.
Object selection occurs in this manner:
The lateral position of the object versus the predicted course of the ACC
system’s own vehicle is calculated.
A calculation is made of the object’s "lane probability", that is, which lane
the object is most likely to be in.
Lane probability is a main input into the next step, that of a "plausibility
attribute". Together with the frequency and reliability of object detection,
this determines the degree of plausibility that the detected vehicle is in the
same lane as the ACC car.
The object is selected as the target only if the degree of plausibility is
sufficient. This plausibility is based only on moving objects – ACC systems
ignore stationary objects when selecting targets.
The first step – that of locating the object relative to the predicted course
of the ACC car – is most critical. This diagram shows three cars travelling
around a curve on a multilane road. Car 3, the car equipped with the ACC, is at
the bottom of the diagram. Without an ability to accurately model the predicted
course of the ACC car, the system would expect to follow Course B and therefore
sense Car 2 as being ahead of it in its lane. However, the ACC car will actually
follow Course A and so must sense Car 1 as being ahead of it.
Course prediction is based on the "trajectory curvature". That is, the change
in direction that the car is undergoing as a function of the distance travelled.
This is determined by sensors detecting steering angle, lateral acceleration,
yaw and the difference in left/right wheel speeds. The effect of crosswinds,
road camber and differences in wheel diameters can all reduce trajectory
curvature prediction. Combining the techniques reduces the probability of
error.
In addition, the ACC system can use the current and past positions of
stationary and moving objects to determine the projected course of the car. This
can be carried out by analysing the apparent lateral movement of vehicles in
front as they enter a bend, and analysing near-road stationary objects.
Special logic is used in sharp bends. If it is sensed that the car is
negotiating a sharp bend, a reduction is made in the maximum permissible
acceleration (note that in this context, acceleration also refers to
deceleration) so as to maintain vehicle stability. Secondly, as described above,
the effective range of the radar beam is much reduced in corners and so the ACC
modifies the allowable acceleration to suit this reduced "visibility". Finally,
if the target car disappears from view, logic prevents the ACC vehicle from
suddenly speeding up.
The Bosch ACC uses the 6-level control sequence shown above. The first level
is the input of data from the radar, wheel-speed sensors, yaw sensor and other
sensors. The second level is to identify the moving objects ahead of the car and
assess their plausibility of being in the same lane. In this step the data from
the other car system sensors is assessed to determine the degree of curvature of
the road.
Once this has been done, the system can calculate the projected trajectory of
the ACC car and track and predict the course of other vehicles. A target vehicle
is established – normally it will be the one calculated as being ahead of the
ACC car in the same lane. However, this is not always the case: if vehicles
ahead of the ACC car (or the ACC car itself) change lanes, a group of several
possible target vehicles can be considered.
The next step is the calculation of the required acceleration. The actuation
system by which the car’s speed is to be changed is selected (it can be
throttle, brakes or transmission) and then finally, this control is exerted.
The driver has control over two functions: the set speed and the distance to
be maintained between the ACC car and the car ahead. As mentioned earlier, the
distance is set by means of a requested time gap which is generally in the range
of 1 – 2 seconds.
On the Road
Most cars equipped with ACC use a similar driver interface. The selected
cruise speed is shown by an illuminated segment or LED on the speedometer. The
selected gap spacing is shown diagrammatically on a dot matrix or TFT display –
for example, by the spacing between two diagrammatic cars. When the ACC is
tracking a car, another symbol illuminates on the dashboard display. In this
way, the requested and actual vehicle speeds, the requested gap and the tracking
action of the ACC can all be quickly and easily seen.
Current ACC systems are suitable for use primarily on freeways and open rural
roads. They will not brake a vehicle to a standstill, even if the vehicle is
aimed straight at a roadside obstacle. Furthermore, if the traffic ahead is
stopped, an alarm may sound but again the vehicle will not be emergency braked.
Such collision avoidance systems are in the pipeline but as was remarked at the
beginning of this story, ACC is only the first step on that road. However, it’s
a pretty impressive step...
Driving with Adaptive Cruise Control
In the past we’ve been able to drive the Audi A8 4.2, a car that features
Bosch Adaptive Cruise Control. And what was it like? In a word, brilliant.
We didn’t have a chance to test it on tight, winding country roads but in
freeway conditions it was superb. Speed selection is available only in 10 km/h
increments – which is fine when you no longer need to ‘tap-up’ and ‘tap-down’ in
tiny increments, trying to maintain a constant gap to the car in front. As you
would expect with a system that maintains a constant time gap, at slow speeds
the Audi would creep up on the car in front and at higher speeds it would drop
back. All automatically, of course.
If the car ahead slowed abruptly, the Audi would automatically apply the
brakes – and if it was deemed by the system to be an emergency stop, an audible
alarm sounded and you were expected to brake. A green symbol showed on the
instrument display when the car in front was within the minimum safe distance –
this changed to red when driver braking was needed.
But describing the system in step by step detail makes it sound more
cumbersome than it really was. This is literally a set-and-forget system – on a
drive from Sydney to Canberra or Melbourne, it would be simply awesome.
If the price of the technology drops as it has for other car innovations,
we’re happy to go on record and say that in the foreseeable future – say, in 10
years time – all cars with cruise control will have a radar proximity function.
It just works so well....
|