PRE-CRASH EXTRACTION OF THE CONSTELLATION OF A FRONTAL COLLISION BETWEEN TWO MOTOR VEHICLES

One of the strategic objectives of the European Commission is to halve the number of road traffic fatalities by 2020. In addition, in 2010, the United Nations General Assembly initiated the "Decade of Action for Road Safety 2011-2020" to reduce the number of fatalities and decrease the number of road traffic injuries. To address the scourge of road traffic accidents, this thesis presents a research study which has devised and evaluated a novel algorithm for extracting the constellation of an unavoidable frontal vehicle-to-vehicle accident. The primary research questions addressed in this work are:
• What are the most significant collision parameters which influence the injury
severity for a frontal collision between two motor vehicles?
• How to extract the constellation of a crash before the accident occurs?
In addition, the secondary research questions given below were addressed:
• How to integrate physical constraints, imposed on the rate of acceleration of a
real vehicle, together with data from vehicle-to-vehicle (V2V) communication,
into the crash constellation extraction algorithm?
• How to integrate uncertainties, associated with the data captured by sensors
of a real vehicle, into a simulation model devised for assessing the performance
of crash constellation extraction algorithms?
Statistical analysis, conducted to determine significant collision parameters, has
identified three significant crash constellation parameters: the point of collision on the vehicle body and the relative velocity between the vehicles; and the vehicle alignment offset (or vehicle overlap). The research reported in this thesis has also produced a novel algorithm for analysing the data captured by vehicle sensors, to extract the constellation of an unavoidable vehicle-to-vehicle frontal accident. The algorithm includes a model of physical constraints on the acceleration of a vehicle,
cast as a gradual rise and eventual saturation of vehicle acceleration, together with the acceleration lag relative to the timing of information received from V2V communication.
In addition, the research has delivered a simulation model to support the evaluation of the performance of crash constellation extraction algorithms, including a technique for integrating (into the simulation model, so that the simulation can approach real-world behaviour) the uncertainties associated with the data captured by the sensors of a real vehicle.
The results of the assessment of the soundness of the simulation model show that the model produces the expected level of estimation errors, when simulation data is considered on its own or when it is compared to data from tests performed with a real vehicle.
Simulation experiments, for the performance evaluation of the crash constellation extraction algorithm, show that the uncertainty associated with the estimated time-to-collision decreases as vehicle velocity increases or as the actual time-to-collision decreases. The results also show that a decreasing time-to-collision leads to a decreasing uncertainty associated with the estimated position of the tracked vehicle,
the estimated collision point on the ego vehicle, and the estimated relative velocity between the two vehicles about to collide. The results of the performance assessment of the crash constellation extraction algorithm also show that V2V information has a beneficial influence on the precision of the constellation extraction, with regards to the predicted time-to-collision, the predicted position and velocity of the oncoming vehicle against which a collision is possible; the predicted relative velocity between the two vehicles about to collide, and the predicted point of collision on the body of the ego vehicle.
It is envisaged that the techniques, developed in the research reported in this thesis, will be used in future integrated safety systems for motor vehicles. They could then strongly impact passenger safety by enabling optimal activation of safety measures to protect the vehicle occupants, as determined from the estimated constellation of the impending crash.