Transportation – Safety

Sensors are now being developed that will help tomorrow's vehicles avoid accidents. The next step is to put these sensors online so that each vehicle becomes part of a real-time, intelligent traffic information network.

On the drawing board in Europe, the U.S. and Japan are cameras, radars, and a range of information systems that may essentially transform our vehicles into mobile sensors designed to warn us whenever certain categories of danger are near. "Tomorrow's cars and trucks," says Professor Anton Anthofer of the University of Applied Sciences at Amberg-Weiden near Regensburg, Germany, "will be loaded with systems that will not only call the driver's attention to potential threats, but will use cell phone and GPS systems to transmit relevant portions of their output, such as information on road conditions, speed and direction, to nodes that will in turn pump the information back to other vehicles as needed." Anthofer should know. A specialist in automotive electronics, he works closely with Siemens to help the company coordinate its development activities with the German government's INVENT (intelligent traffic use technology) program, which was initiated in July of 2001. The 76 mill. € ($66 million) research program, which includes many of Germany's biggest companies, is divided into three major areas: driver assistance and active safety, traffic management, and traffic management and transport logistics. Its goal is to reduce accidents and congestion by making traffic itself intelligent.

According to Martin Hummel, a specialist in automotive safety electronics, the cameras will be able to look ahead about 60 m at speeds below 50 km/h and alter the vehicle's speed dynamically even as different vehicles pop in and out of the lane ahead of it. "In heavy traffic, the car could feel as though it's practically driving itself," says Hummel. "The idea is to provide an instantaneous reaction to rapidly slowing traffic and thereby reduce the incidence and severity of rear-end collisions while also reducing driver stress." Statistics from the U.S. National Highway Transportation Safety Administration (NHTSA) show that rear-end collisions account for over 1.5 million accidents per year.

New data compression algorithms could allow such camera mounted systems to operate at all driving speeds. For instance, at Roke Manor Research (RMR), a U.K.-based business owned by Siemens, scientists have come up with a process called "structure from motion" that sees key feature points in successive video frames and continuously triangulates their distance relative to itself. "If a vehicle is coming up behind you or beside you, the system can compute its time to impact or likely point of closest approach. In this sense, vision can be used as an anti-collision system," says Chris Gilham of the RMR Sensors Business Unit, which is developing the system. "All of this is done in real time by reducing the essential image data to only around 400 or 500 feature points in each image. That's what makes it so fast," he explains.

In addition to providing dynamic cruise control, cameras can play a decisive role in helping to avoid accidents in which vehicles stray into adjoining lanes or run off the road as a result of driver error or exhaustion—two causes of accidents that account for up to 1.4 million crashes per year in the U.S. alone. "We have developed algorithms that allow the cameras to track lane markings and road edges," says Dr. Thorsten Köhler, a specialist in video sensor systems at Siemens VDO. "Any departure from the lane not preceded by a blinker signal would trigger a warning sound or other stimulus designed to alert the driver."

In addition to cameras, Stop-and-Go Assistant will probably also use bumper-mounted radar heads. Radar's outstanding precision in measuring distance and its imperviousness to weather, dirt and lighting conditions, make it a valuable adjunct to visual sensors. But demand for long-range (150 m, 77 GHz) radar systems that are applicable to adaptive cruise control (ACC) has been limited to a handful of high-end cars in Europe.

Siemens VDO engineers are therefore concentrating on short-range 24 GHz radar sensing. "Manufacturers have voiced considerable interest in this technology because it has a variety of applications," explains Dr. Reiner Doerfler, Manager, Radar Sensor Systems, at Siemens VDO. "This technology can sense objects around the car at a distance of 3 to 20 m, which means that it could be used not only for stop-and-go adaptive cruise control, but also for pre-crash sensing, that is, to prime the airbags a split second before a crash, for blind spot detection, and, when combined with cameras, pedestrian detection and automatic parking.

In the U.S., where European-style curved mirrors are not permitted, manufacturers are particularly interested in blind spot protection. Siemens VDO's position in this potentially enormous market is strong. "Our system is small enough to be built into the housings of the outer mirrors, which is the ideal location for it," says Doerfler.

As researchers look ahead and see a proliferation of cameras and radar sensors on tomorrow's cars, they are asking themselves how the resulting flood of data can be harmonized and how such systems can be made affordable. "Right now, we are trying to identify the level at which information between radar and video sensors can be fused," says Doerfler. "Cameras have excellent lateral resolution; radars are poor in this area. On the other hand, radars are good at measuring distances and differential speeds." One step appears to be clear: fewer radar heads will be needed if they can be synchronized. In the context of a European Union research project called Radar Net, Siemens VDO and others are therefore studying what the optimal frequencies would be for a connection between the heads. "The ultimate goal," says Doerfler, "is to have a 360 ° high-resolution view around the vehicle."

Expected to debut in two years, the platform will make it possible, for instance, for your car system to recognize your airbag sensors and give their signals top priority if they are activated. It will also allow the navigation system to wirelessly download addresses from your PDA, cell phone or computer and plot a course to your stated destination. "Our idea is to turn cars into open systems," says Dr. Hongjun Pu, Project Manager for the INVENT Traffic Management project at Siemens VDO. "Once you have an open multimedia platform using a standard application programming interface, if you add something like new external cameras, you will be able to download the software into the car using your cell phone. The software will then be integrated with existing services."

Starting in 2003, when TLA is expected to hit the market, something entirely new will happen: cars and trucks will start to become part of a telematic chain. Their onboard sensors will not only help to guide drivers out of harm's way, but will become the eyes and ears of a comprehensive, real-time travel information system that will become increasingly detailed with each new vehicle that hits the road.

Crucial information about road, traffic and weather conditions derived from sensors will flash through each vehicle's navigation system for geographical postmarking before being zipped to a service provider for processing. The same information—possibly combined with information on other events that could affect traffic—will be broadcast. Each vehicle that receives these messages will channel the data through its navigation system to determine whether the information applies to its "flight plan" and, if so, will suggest course or speed alterations to its driver.

The value of such a system is potentially immense. According to Germany's national automobile association, traffic jams cost the nation's economy at least 250 mill. € ($217 million) a day. And in the U.S., the Department of Transportation (DOT) estimates that congestion costs over $48 billion a year in lost productivity. But the real tragedy is that accidents cause millions of injuries and thousands of fatalities per year. Sensor systems are expected to put a major dent in these statistics. The DOT, for instance, estimates that advanced lane-keeping and collision-avoidance technologies could prevent 71,000 crashes a year in the U.S., and that other sensor systems addressing rear-end and roadway departure crashes could prevent an additional 1.1 million crashes per year.

"There's no doubt that sensor-based crash avoidance technologies have the potential for saving thousands of lives per year," says Prof. Anthofer. "The real question now is how quickly can we get from here to there."

Arthur F. Pease