A methodology for the analysis of in-vehicle operating data and design of intelligent vehicle systems for improved automotive safety.

摘要

Every year global motorization increases as more motor vehicles are manufactured, and the total number of vehicle miles traveled rises. These increased travel opportunities result in higher numbers of injuries, fatalities, and monetary losses associated with traffic-related crashes. In the last decade, hundreds of thousands of people were killed by vehicle collisions in the United States. The World Health Organization has labeled traffic crashes as the ninth leading cause of global disease; by the year 2020, traffic crashes are expected to rise to number three. An opportunity exists to improve global human safety through research and innovation in driver training and evaluation and advanced vehicle safety systems. In this dissertation, four research studies were conducted: creation and evaluation of a safe driving program, driver classification using in-vehicle data collection and analysis, development of an obstacle avoidance warning system, and design of a run-off-the-road recovery controller.;The most critical component of vehicle safety is the driver. For this reason, a safe driving program was developed to improve driver skills, knowledge, attitudes, and behaviors. The program consisted of driving and tent modules that were targeted to younger and less experienced drivers. Standardization of the modules allowed for student assessment using subjective and objective evaluation tools. A total of 86 students participated in a case study. Comparison of pre- and post-event tests indicated a 10% net increase of knowledge with a student and parent satisfaction level of 89.6%. One driving module focused on a tailgating scenario using a custom apparatus to simulate a tailgating situation. For this module, 75% of the evaluated students received a passing grade (85% or above), while the other 25% received valuable feedback on their specific driving eficiencies.;The evaluation of normal driving tasks can be used as a tool to supply drivers with feedback regarding inadequate skills or poor behaviors, while providing off-line users with risk assessment. Three custom analysis techniques were developed to analyze real-world driver behavior and provide a normalized driving score, ultimately creating a driver classification system and risk assessment. A five-person case study was performed to demonstrate the capability of the developed methodologies; the results successfully differentiated each driver's overall performance.;Driver safety may also be improved through the use of advanced on-board vehicle safety systems. A customizable hardware-in-the-loop steering simulator was used to create an obstacle avoidance system. Variable levels of vibration were provided to the driver through the steering wheel to communicate critical roadway information. Laboratory results demonstrated that haptic steering feedback improved driver performance as measured by a 62% reduction in obstacle hit rates. In addition, small reductions were found in peak steering wheel angle and peak vehicle yaw rate, as well as a 10m (32.8ft) increase in the reaction distance to the obstacles.;For situations involving a run-off-the-road scenario, a more invasive autonomous vehicle system may provide a greater safety benefit by removing driver error from the recovery process. Two steering and braking controllers, Sliding Mode and State Flow, were designed and simulated using the CarSim and Matlab/Simulink software packages. The complete simulation results illustrated that these controllers outperformed the driver steering model by safely performing the recovery process over a range of vehicle and roadway conditions. Peak lateral error was reduced by 447% and 663% for the Sliding Mode and State Flow controllers, respectively. In addition, the controllers' performances were greatly influenced by the vehicle speed and roadway surface friction.;This research study proposes a multi-phased approach to improve driver safety. Future opportunities for driver improvement are highlighted by further development of training modules, increasing the number of events, and a large-scale dissemination of the driver classification system. Concurrently, further exploration of the human-vehicle interface will improve the haptic feedback warning system. Lastly, a better understanding of the vehicle/road interface coupled with robust vehicle parameter estimators will advance the performance of autonomous vehicle controllers.