Based on cooperative sensing technology, the pedestrian protection system detects, identifies, locates and tracks pedestrians. It is specially designed to deal with situations where there is no line of sight. Vulnerable road users such as pedestrians or cyclists are equipped with active transmitters. These can be embedded in clothes, satchels or cell phones.
The tracking technology incorporated into the vehicle uses a multiple-antenna system to identify pedestrians and cyclists based on active transmitters worn or carried by them. If a pedestrian is close to the car, they will be reliably identified as a vulnerable road user. The system then determines how far away the mobile transmitter (and hence the pedestrian) is and in which direction. If there is a risk of collision, the system will either warn the driver or take action to avoid or mitigate it. This is possible even if the pedestrian is obscured by parked cars or buses.
The pedestrian protection system integrated into Ko-TAG starts by tracking all pedestrians in its sensing range who are equipped with a transponder system. A multiple-antenna system incorporated into the vehicle measures the distance to each object tracked and the angle it forms with the vehicle. The tracking system will stay connected to a transmitter even if there is no line of sight.
Apart from on-board angle and distance measurement, the tracking system also uses an inertial measurement unit, which is integrated into the transceiver and acquires data on the pedestrian's speed and direction. The information collected is fed into the tracking process, helping refine results.
Another advantage of the integrated inertial measurement unit is that it allows the transponder system to be deactivated whenever it has been stationary for a long period of time. This means that the transceiver unit can be switched off to minimize power consumption.
The ability to track pedestrians obscured from view provides a basis for adding further capability through data fusion, e.g. using input from video-based image processing. Since the tracked path of the pedestrian is sufficient to predict exactly where they will step into the road even if they cannot yet be seen, an area of interest can be defined within the image and specially processed. This area can then be closely scanned for shapes or even partial shapes (e.g. legs, chest or head) of pedestrians.
Ko-TAG can thus generate valuable additional information for image-based pedestrian detection. Combining the two approaches will ensure reliable identification of pedestrians and yield very high quality data on their lateral and longitudinal distance. This will form the basis for a range of protective measures, including an automatic emergency stop.
Once the system has detected a pedestrian and tracking has yielded position and velocity information, the risk of colliding with the pedestrian can be calculated. This is a complex process because both the pedestrian and the driver may act in a variety of ways, which means a collision is impossible to predict with absolute certainty until it is imminent.
As a rule, however, the protection system will need to take action much earlier if it is to be able to limit the severity of a potential accident. This requires the ability to accurately identify and assess the pedestrian's range of possible behaviors, as pedestrians are more maneuverable than vehicles and can change direction very quickly.
The pedestrian's behavior is predicted with the aid of a statistically based behavior model. This method involves first deriving a model of the pedestrian's behavior from previous observation of a large number of pedestrians and then determining his or her most likely behavior in a given situation.
Based on the general mechanics of human locomotion and on observed behavior, it is possible to identify patterns that typically precede specific changes in a person's movements, indicating, for instance, that they are likely to step into the road. In the case of Ko-TAG, a combination of early indicators derived by tracking the pedestrian's movements and information from the inertial measurement unit is used to predict a hazardous change in the pedestrian's behavior.
If the collision probability exceeds a threshold, the amount of time remaining until contact with the pedestrian (also known as time to collision or TTC) is calculated on board the car. Depending on how much time remains as well as on the speed of the vehicle and the specific situation encountered, the planned demonstration system will initiate a particular type of intervention.
If the time to collision exceeds 1.5 seconds, the driver will still be able to respond effectively to a warning. Additionally, the system can prepare the vehicle for an emergency stop by precharging the brakes and optimizing brake force distribution.
If a collision is very likely to be unavoidable, the system can initiate an emergency stop. In addition to actuating the brakes, it can optimally prepare the vehicle's occupants for the emergency stop by activating the belt pretensioners and head rests (source: http://ko-fas.de/).