Track Update. Tracks are updated from the associated measurements. A number of tracks are updated to ensure that all relevant tracks are initiated and subsequently maintained. Supporting tasks are used to initiate new tracks and terminate old tracks with scoring and prioritizing measures derived for the lane change support functions. In order for this sensor fusion to be embedded and run in real time, TRW Conekt has developed a high-performance fusion processing platform. High-performance was necessary in this case as 18 tracks were to be maintained with up to 24 sensor target measurements every 40 ms. To achieve greater accuracy, each track was computed using a Kalman filter. (A Kalman filter is a means of estimating information from indirectly related or indistinct “noisy” measurements and is often applied to tracking problems). Figure 3 shows a view of the lanes surrounding the host vehicle with target returns from each of the radars and the computed tracks shown. The track history of one vehicle overtaking the host is highlighted. The track data is used for lane change support where information is provided to the driver, potentially as an audible communication through the entertainment system or as a visual signal through illuminated icons in wing (side) mirrors or the instrument cluster. Intersection Assistance British accident statistics show that 36.1% of fatal accidents are at or within 20 m (66 feet) of intersections.3 Japanese accident statistics show that 45.5% of fatal accidents are at or near intersections.4 U.S. accident statistics show that 22.5% of fatal accidents are at or related to intersections.5 Motivated by such statistics, INTERSAFE, one of the sub-projects funded by the European Union's PReVENT project, set the goal of reducing and ultimately eliminating fatal collisions at EU intersections (see sidebar).

One of the challenges facing intersection safety systems is to accurately position or localize

-------------------------------------------------------------------
4 Statistics 2004 Road Accidents, Japan Traffic Bureau, National Police Agency.
5 Traffic Safety Facts 2004, U.S. Department of Transportation.

a vehicle at an intersection. Global Positioning System (GPS) receivers are obvious candidates, but in built-up areas the measurement accuracy and availability can be poor. In response, TRW Conekt has developed video intersection localization, building on its Video Lane Detection technology experience (see Figure 4).

For lane detection, only parallel lane markings are detected. For intersection localization, a number of lines (including their start and end) need to be detected and associated with lines on a map of the intersection. The map and associated data are used to update the position and motion of the vehicle, in this case using an extended or non-linear form of the Kalman filter. The Kalman filter tracking is enhanced by fusing information from other readily available sensors such as wheel speed and yaw rate sensors. The localization of the vehicle and path prediction of other road users can further be fused by the vehicle systems with other available information, for example, traffic signal status. A picture of the environment surrounding the vehicle is built up, allowing vehicle systems to assess the risks of potentially hazardous situations and so take appropriate action, such as warning the driver.

Autonomous Vehicles
Foresight Vehicle is the UK's prime knowledge transfer network for the automotive industry involving collaboration among industry, academia and government6. Through its involvement in the Foresight Vehicle program, TRW Conekt has coordinated the AutoTaxi project that has developed a safety-critical sensor system for autonomous vehicles7 (see opening picture to this article). This technology is currently aimed at vehicles operating on a network of dedicated guide-ways, a simpler operating environment for the development of safe fully autonomous driving than the road network. Fault-tolerant sensor fusion of radar, video and optical sensors provides information that enables collision and obstacle avoidance and automatic guidance strategies (including closed loop control of braking and steering systems).

Cooperative Sensing
The technology detailed thus far has been concerned with advances in intelligent vehicles through sensor fusion. TRW is a partner in the European Union (EU) program called Technologies for Road Advanced Co-operative Knowledge Sharing Sensors (TRACKSS)8. TRACKSS is part of the EU Sixth Framework program that is the EU's main instrument for research funding in Europe. TRACKSS is looking at the integration of intelligent vehicles and


3 Road Casualties Great Britain: 2004 Annual Report, Department of Transport.
6 For more on the Foresight Vehicle, visit http://www.foresightvehicle.org.uk/
7 For more on the AutoTaxi project, visit http://www.auto-taxi.co.uk/
8 For more on the TRACKSS program, visit http://www.trackss.net