Search Results

For at least 15 years it has been recognized that pre-crash data captured by event data recorders might help illuminate the actions of drivers prior to crashes. In left-turning crashes where pre-crash data are available from both vehicles it should be possible to estimate features such as the location and speed of the opposing vehicle at the time of turn initiation and the reaction time of the opposing driver. Difficulties arise however from measurement errors in pre-crash data and because the EDR data from the two vehicles are not synchronized so the resulting uncertainties should be accounted for. This paper describes a method for accomplishing this using Markov Chain Monte Carlo computation. First, planar impact methods are used to estimate the speeds at impact of the involved vehicles. Next, the impact speeds and pre-crash EDR data are used to reconstruct the vehicles’ trajectories during approximately 5 seconds preceding the crash.

A continuing topic of interest is how to best use information from Event Data Recorders (EDR) to reconstruct crashes. If one has a model which can predict EDR data from values of the target variables of interest, such as vehicle speeds at impact, then in principle one can invert this model to estimate the target values from EDR measurements. In practice though this can require solving a system of nonlinear equations and a reasonably flexible method for carrying this out involves replacing the inverse problem with nonlinear least-squares (NLS) minimization. NLS has been successfully applied to two-vehicle planar impact crashes in order to estimate impact speeds from different combinations of EDR, crush, and exit angle measurements, but an open question is how to assess the uncertainty associated with these estimates. This paper describes how Markov Chain Monte Carlo (MCMC) simulation can be used to quantify uncertainty in planar impact crashes.

Martinez and Schlueter [6] described a three-phase model for reconstructing tripped rollover crashes, where the vehicle's path is divided into pre-trip, trip, and post-trip phases. Brach and Brach [9] also described this model and noted that the trajectory segmentation method for the pre-trip phase needed further validation. When a vehicle leaves a measurable yaw mark at the start of its pre-trip phase it might be possible to compare estimates from the three-phase model to those obtained using the critical speed method, and this paper describes Bayesian reconstruction of two such cases. For the first, the 95 percent confidence interval for the case vehicle's initial speed, estimated using the critical speed method, was (64 mph, 81 mph) while the 95 percent confidence interval via the three-phase model was (66 mph, 79 mph).

The critical speed method uses measurements of the radii of yawmarks left by vehicles, together with values for centripetal acceleration, to estimate the speeds of the vehicles when the yawmarks were made. Several field studies have indicated that equating the centripetal force with braking friction produced biased estimates, but that the biases tended to be small (e.g. within 10%-15% on average) and led to underestimates, suggesting that the method can be useful for forensic purposes. Other studies, however, have challenged this conclusion. The critical speed method has also seen use in safety-related research, where it is important to have a reliable assessment of the uncertainty associated with a speed estimate. This paper describes a variant of the critical speed method, where data from field tests lead to an informative prior probability distribution for the centripetal acceleration.

In statistical modeling, cross-validation refers to the practice of fitting a model with part of the available data, and then using predictions of the unused data to test and improve the fitted model. In accident reconstruction, cross-validation is possible when two different measurements can be used to estimate the same accident feature, such as when measured skidmark length and pedestrian throw distance each provide an estimate of impact speed. In this case a Bayesian cross-validation can be carried out by (1) using one measurement and Bayes theorem to compute a posterior distribution for the impact speed, (2) using this posterior distribution to compute a predictive distribution for the second measurement, and then (3) comparing the actual second measurement to this predictive distribution. An actual measurement falling in an extreme tail of the predictive distribution suggests a weakness in the assumptions governing the reconstruction.

This project team conducted a life-cycle-based environmental evaluation of new, lightweight materials (e.g., titanium, magnesium) used in two concept 3XVs -- i.e., automobiles that are three times more fuel efficient than today's automobiles -- that are being designed and developed in support of the Partnership for a New Generation of Vehicles (PNGV) program. The two concept vehicles studied were the DaimlerChrysler ESX2 and the Ford P2000. Data for this research were drawn from a wide range of sources, including: the two automobile manufacturers; automobile industry reports; government and proprietary databases; past life-cycle assessments; interviews with industry experts; and models.