ARTICLE MAY 2003 DRIVER-ASSISTANCE SYSTEMS FOR SAFETY AND CONVENIENCE On average around the world, someone is killed in or due to a vehicle accident almost every minute of every day. The economic damage of such accidents has been estimated at over US$ 600 billion a year for Europe, Japan and the USA. Electronics have to play a major role in improving global automotive safety, particularly by timely recognition of a threat and appropriate manoeuvres for collision avoidance ? either by providing a suitable warning to the driver or even by automatic control of the vehicle itself. Both approaches require appropriate sensors with an effective sensor vision system and a matched human-machine interface. Improving driver vision Driver-assistance systems can help to cut accidents. In particular, vision improvement systems for night driving can reduce the number of collisions ? 40% of all fatal accidents occur in darkness although only 20% of driving is done at night. And a US study reveals that in more than 60% of all fatal accidents pedestrians are involved. Apart from weather conditions, the lights of oncoming vehicles are the main cause of limited vision at night. Adaptive curve light was introduced in the 1960s and intelligent headlight systems, adapting the beams of the headlights to specific traffic situations, are under development. A far infra-red (FIR) night vision improvement system is offered as an option on some US cars, and has achieved good market acceptance. The sensor detects the FIR image of objects in front of the vehicle and shows them as a video picture projected on the windscreen as a head-up display (HUD). In Europe, a near infra-red (NIR) system is currently under development. It uses NIR from a modified halogen headlight. The image is detected with a video camera using a new non-linear imager with a high dynamic range. The advantage of this combination is better performance compared with the FIR system because road limit lines are seen clearly at a wide range and thus a differentiation of objects within or outside the line can be made. In addition, the video camera used for night vision improvement offers a huge potential for other vision-based safety and comfort functions at night or during the day ? such as lane departure warning, traffic-sign recognition, and adaptive cruise control (ACC). An analysis of the causes of accidents outside urban areas in Germany indicates that more than one-third can be traced back to a lane change or unintentionally drifting from a lane. Sensing systems can provide assistance by monitoring a driver?s blind spot and checking that the vehicle is staying in its lane. Another area in which such systems can provide assistance involves front-end collisions. These situations can be avoided through crash warning systems and active brake intervention ? ACC is the first step in this direction. Electronic surround sensing is the basis for numerous driver-assistance systems ? systems that warn or actively intervene. Due to limited availability of sensors, only a few driver-assistance systems have become established. Typical is the monitoring of close objects using ultrasound technology. Sensors can be integrated in the front bumper to provide an acoustic or optical warning to the driver when approaching an obstacle. Such systems are already in mass production in many vehicle models in Europe. Various sensor technologies A wide range of applications is possible with driver-assistance systems, based on various sensor technologies. These can be subdivided into active systems with automatic vehicle intervention, and passive systems for driver information and passenger protection. The functions of active safety systems range from a simple parking brake, which automatically slows a vehicle before reaching an obstacle, to computer-supported control of complex driving manoeuvres to avoid collisions. For example, the automatic emergency braking feature intervenes if a crash is unavoidable. In its highest levels of refinement, active driver-assistance systems intervene in steering, braking and engine management to avoid colliding with an obstacle. ACC belongs to the group of active comfort systems. If longitudinal guidance is augmented by lane-keeping assistance with a video-based system for lateral guidance, automatic driving is possible in principle. Driver-support systems without active vehicle interaction can be viewed as a pre-stage to vehicle guidance. They only warn the driver or suggest a driving manoeuvre. One example is a ?parking assistant?. This gives the driver steering recommendations when parking in order to fit optimally in previously automatically determined space. Secondary surveillance radar (SRR) sensors determine objects in the blind spot: the driver receives optical instructions via an illuminated diode near his outside mirror and an acoustic warning. Standing obstacles can be filtered out so the driver is not overwhelmed with signals. Microwave devices starting from scratch Microwave devices above 50 GHz are starting to be used in automotive applications, particularly for ACC. However, useful automotive designs had to be deployed from scratch, based on reasonable cost, high environmental robustness and small size. Such a unit can be divided into three parts: a radar transmitter/receiver (transceiver), including analogue signal processing; a digital signal processing unit; and a control unit. A Gunn-oscillator on the radar transceiver delivers output power for a number of radar beams and the same number of mixers for the demodulation of the received waves. Modulation is controlled by a frequency-locked-loop using a reference oscillator. This control circuit guarantees linearity of the modulation sweeps and compliance of the frequency band limits. Digital signal processing for ACC needs number-crunching chips with the highest calculation power in the car. A European commercial radar sensor uses a 24-bit digital signal processor assisted by an application-specific IC (ASIC) for fast Fourier transformation. The control unit is based on a 16-bit microcontroller linked to the actuator subsystems and to external connections for operation switches. It calculates the control algorithms and the basic part of the self-diagnostics. Long- and short-range detection A long-range detection system based on a 77-GHz radar is already used for ACC in BMW and Fiat cars. With a distance range from 2 to 120 m, it is able to detect multiple objects and to measure distance, relative speed and the angle simultaneously. The angle measurement is based on three beams, which are working in parallel without any switch. The raw radar data is tracked and filtered so that the ACC control algorithm can select the relevant object. Information from the ACC is used either to warn the driver if too close to a vehicle in front or to maintain automatically a safe distance from the vehicle ahead. The driver indicates the desired speed and safety margin. The system maintains these specifications through independent interventions in brake and engine management. In current versions, the system works at speeds over 30 km/h. In the absence of lateral resolution for this radar device, stationary objects have to be suppressed to avoid too many false alarms, decreasing acceptance of the system. Short-range microwave sensors are also being introduced in series production. An existing first sensor generation has a set-up with discrete radio-frequency (RF) components. The microcontroller and signal processing are integrated on the circuit board. Using a 24-GHz radar with a pulse width of 300 ps, the measuring cycle is 10 ms. With a multi-target capable sensor and radiated power far below 1 mW (mean), the detection range is between 0.25 and 20 m with an accuracy of ? 2 cm. The beam angle is ? 45? horizontal and ? 15? vertical respectively. Dielectric lenses can be used in front of the aerial to vary these angles. The major benefit to the customer is the multi-usage capability of this radar-sensor ? opening up the opportunity to build a sensor platform. It?s introduction is currently limited by the pending release of the frequency band for these applications. This release has been given for the USA, and is expected to be granted soon in Europe. As a first step, short-range 24-GHz microwave sensors will be introduced. These can be used to build a virtual safety belt around the vehicle, covering a variety of functions. Although the introduction of this type of sensor will most probably be driven by one or two individual functions, studies have shown that a minimum of eight sensors (depending on car size) allow an almost complete surround view with all its functionality. This multi-usage of sensors would also reduce the cost of the system. Such sensors allow general distance and velocity measurement with high resolution, providing the opportunity for:
Screen-mounted video sensors Video sensors can also play an important role, mounted on a windscreen adapter. Since the brightness of the scene cannot be controlled in the automotive environment, the dynamic range of common CCD (Charge-Coupled Device) technology is insufficient and high dynamic range imagers are needed. Such a sensor in a vehicle would make use of CMOS technology with non-linear luminance conversion to cover a wide luminance dynamic range and so significantly outperform current CCD cameras. Video technology will first be introduced for comfort functions that provide a clearer view for intervention by the driver. For example, a rear-view camera with appropriate display facilities is envisaged that improves a driver?s view when reversing. More sophisticated functions necessitate new imager technologies that provide a sufficient luminance dynamic range to cope with all illumination situations, as mentioned earlier. The enormous potential of video sensing is obvious from the performance of human visual senses. Although computerised vision has not, until now, achieved similar performance, a respectable amount of information and related functions can readily be achieved by video sensing, including:
Rapid pace of development Sensors to detect the vehicle environment are being developed at a rapid pace. New functions are quickly integrated because of their importance for safety and comfort. Meanwhile, complexity of the functions will grow steadily. The first significant step will be made with the introduction of a 24-GHz platform fulfilling a multi-use concept for sensors. The second stage will follow a couple of years later when new CMOS imagers will be available. A significant extension of functions is then expected, and within three to four years, when sensor chips with high dynamic range will become available for large series production, versatile video sensors will be widely used in cars. New technical applications are continuing to become available to improve driving comfort and safety. Both already anticipated uses and future new functions will require improved silicon features to provide powerful new sensors, greater computing power and bigger memories to handle the signal processing and higher levels of intelligence needed. The second phase of the MEDEA+ programme is opening the door to such uses and relevant supporting technology ? particularly the integration of such applications into complete system on chip (SoC) devices that are sufficiently robust and reliable to operate in the harsh environment of the motor vehicle. | |||
