Abstract:

In this dissertation, first the concept of Cooperative Vehicle Safety (CVS) is introduced and explained. The CVS concept provides warnings or situation awareness displays to drivers based on information about the motions of neighboring vehicles obtained by wireless communications from those vehicles, without use of any ranging sensors. This has the advantage of being a potentially inexpensive set of onboard vehicle equipment (compared to ranging sensors that could provide 360 degree coverage), as well as providing information from vehicles that may be occluded from direct line of sight to the approaching vehicle.

We present the architecture of the system. The architecture consists of four major sub-systems: (1) "Estimator": that estimates the states of the vehicle. (2) "Neighboring Vehicle Map": that is responsible for broadcasting messages to the neighboring vehicles as well serving information about the neighboring vehicles. (3) "Safety Applications": that receive information from the above two components, (4) "In-Vehicle Display": that is the interface to the driver.

In this dissertation, we present novel solutions for sub-systems 1 and 2 in order to enable CVS.

With respect to the "Estimator", we design an extended Kalman filter for vehicle positioning with lane-level accuracy. We present data from approximately 60 kilometers of driving in urban environments including stops, intersection turns, U-turns, and lane changes, at both low and high speeds. The data show that the filter estimates position, speed, and heading with the accuracies required by CVS.

The estimator design was implemented in the CVS prototype built by the University of California Partners for Advanced Transit and Highways (PATH) and General Motors Research and Development in 2004. The prototype consisted of five vehicles equipped with CVS technology. We describe the prototype and highlight the main technical challenges to be overcome for large scale implementation. This dissertation addresses the challenges.

With respect to the "Neighboring Vehicle Map", we first study the pre-crash geometries of various crashes that might be ameliorated by CVS. We perform the crash analysis in order to find the set of requirements on communications in between cars. We categorize all types of driving environments into two major environments: Intersection and Non-Intersection. For each case, we identify the minimum set of neighbors for vehicular communication.

Next, we design the vehicular data exchange protocol for CVS. The protocol specifies: content of the communication messages, communication policy, and communication range. The communication policy is based on threshold crossing policy. The communication range alternates between a maximum pre-defined range (e.g. 500 meters) and an adaptive range.

Finally, we investigate performances of several communication schemes for position tracking of neighboring vehicles via large scale traffic and network simulation. We use application level reliability performance metrics. Our performance measure metrics include communication losses.