Search

Drivers’ take-over reaction time in partially automated vehicles is a fundamental component of automated vehicle design requirements, and take-over reaction time is affected by many factors such as distraction and drivers’ secondary tasks. This study built cognitive architecture models to simulate drivers’ take-over reaction time in different secondary task conditions. Models were built using the queueing network-adaptive control of thought rational (QN-ACTR) cognitive architecture. Drivers’ task-specific skills and knowledge were programmed as production rules. A driving simulator program was connected to the models to produce prediction of reaction time. Model results were compared with human results in both single-task and multi-task conditions. The models were built without adjusting any parameter to fit the human data. The models could produce simulation results of take-over reaction time similar to the human results in take-over conditions with visual or auditory concurrent tasks, as well as emergency response time in a manual driving condition. Overall, R square was 0.96, root mean square error (RMSE) was 0.5 s, and mean absolute percentage error (MAPE) was 9%. The models could produce simulation results of reaction time similar to the human results from different task conditions. The production rules are plausible representations of drivers’ strategies and skills. The models provide a useful tool for the evaluation of take-over alert design and the prediction of driver performance.

Aggressive driving has attracted significant attention recently with the increase in related road traffic collisions occurring in China. This study aims to investigate the effect of driving skills on aggressive driving behaviors and traffic accidents to find implications for traffic safety improvement in China. A total of 735 Chinese drivers were recruited to complete a self-reported survey including demographic information, the translated Driver Skill Inventory (DSI), and Driver Aggression Indicator Scale (DAIS). Exploratory factor analysis was first conducted to investigate the factor structures of DSI and DAIS among Chinese drivers. Unlike the two-factor solution (i.e., perceptual-motor and safety skills) found in other studies, the current study result revealed a three-factor solution (i.e., perceptual-motor, safety, and emotional control skills) of DSI. Then, the interaction between DSI factors on DAIS factors, demographic variables, and the number of self-reported traffic accidents and offenses was tested by using moderated regression methods. The results revealed the interaction between perceptual-motor skills and safety skills on aggressive warnings committed by drivers themselves. The interactive effect between safety skills and emotional control skills on perceived aggressive warnings was also found. The results suggested that higher ratings of safety skills are essential for buffering the effect of high-level perceptual-motor skills and emotional control skills on aggressive driving in China. In conclusion, policy makers should be interested in understanding the effect of Chinese drivers’ skills on the aggression drivers committed and conceived in traffic. Successful intervention strategies should include all skill factors in the driver training contents.

In past years, the task of automatic vehicle trajectory analysis in video surveillance systems has gained increasing attention in the research community. Vehicle trajectory analysis can identify normal and abnormal vehicle motion patterns and is useful for traffic management. Although some analysis methods of vehicle trajectory have been developed, the application of these methods is still limited in practice. In this study, a novel adaptive vehicle trajectory classification method via sparse reconstruction and mutual information analysis based on video surveillance systems was proposed. The l0-norm minimization of sparse reconstruction in the method was relaxed to the lp-norm minimization (0 < p< 1). In addition, to consider the nonlinear correlation between the test trajectory and the dictionary, mutual information between the test trajectory and the reconstructed one was taken into account. A hybrid orthogonal matching pursuit–Newton method (HON) was developed to effectively find the sparse solutions for trajectory classification. Two real-world data sets (including the stop sign data set and straight data set) were used in the experiments to validate the performance and effectiveness of the proposed method. Experimental results show that the trajectory classification accuracy is significantly improved by the proposed method compared with most well-known classifiers, namely, NB, k–nearest neighbor, support vector machine, and typical extant sparse reconstruction methods.

The development of autonomous vehicles provides effective solutions and opportunities for reducing the probability of traffic accidents. However, because of technical limitations and economic and social challenges, achieving fully autonomous driving is a long-term endeavor. One principal research question is how to choose the suitable driving mode of an intelligent vehicle during stressful traffic events. For this purpose, an on-road experiment with 22 drivers was conducted in Wuhan, China; multisensor data were collected from the driver, the vehicle, the road, and the environment. Driving modes were classified into three categories on the basis of the driver’s self-reported records, and two physiological indexes that use the k-means cluster method were adopted to calibrate the self-reported driving modes. A feature-ranking algorithm based on the information gained was adopted to identify significant factors, and a driving mode decision-making model was established with the multiclass support vector machine algorithm. The results indicated that the SD of the front wheel angle, driver experience, vehicle speed, headway time, and acceleration had significant effects on the driving mode decision making. The driving mode decision-making model demonstrated a high predictive power with a prediction accuracy of 0.888 and area under the curve values of 0.918, 0.91, and 0.929 for the receiver operating characteristic curves. The conclusions provide theoretical support for decision making by the controller of a semiautomated vehicle.

In past years, the task of automatic vehicle trajectory analysis in video surveillance systems has gained increasing attention in the research community. Vehicle trajectory analysis can identify normal and abnormal vehicle motion patterns and is useful for traffic management. Although some analysis methods of vehicle trajectory have been developed, the application of these methods is still limited in practice. In this study, a novel adaptive vehicle trajectory classification method via sparse reconstruction and mutual information analysis based on video surveillance systems was proposed. The l0-norm minimization of sparse reconstruction in the method was relaxed to the lp-norm minimization (0 < p< 1). In addition, to consider the nonlinear correlation between the test trajectory and the dictionary, mutual information between the test trajectory and the reconstructed one was taken into account. A hybrid orthogonal matching pursuit–Newton method (HON) was developed to effectively find the sparse solutions for trajectory classification. Two real-world data sets (including the stop sign data set and straight data set) were used in the experiments to validate the performance and effectiveness of the proposed method. Experimental results show that the trajectory classification accuracy is significantly improved by the proposed method compared with most well-known classifiers, namely, NB, k–nearest neighbor, support vector machine, and typical extant sparse reconstruction methods.

The development of autonomous vehicles provides effective solutions and opportunities for reducing the probability of traffic accidents. However, because of technical limitations and economic and social challenges, achieving fully autonomous driving is a long-term endeavor. One principal research question is how to choose the suitable driving mode of an intelligent vehicle during stressful traffic events. For this purpose, an on-road experiment with 22 drivers was conducted in Wuhan, China; multisensor data were collected from the driver, the vehicle, the road, and the environment. Driving modes were classified into three categories on the basis of the driver’s self-reported records, and two physiological indexes that use the k-means cluster method were adopted to calibrate the self-reported driving modes. A feature-ranking algorithm based on the information gained was adopted to identify significant factors, and a driving mode decision-making model was established with the multiclass support vector machine algorithm. The results indicated that the SD of the front wheel angle, driver experience, vehicle speed, headway time, and acceleration had significant effects on the driving mode decision making. The driving mode decision-making model demonstrated a high predictive power with a prediction accuracy of 0.888 and area under the curve values of 0.918, 0.91, and 0.929 for the receiver operating characteristic curves. The conclusions provide theoretical support for decision making by the controller of a semiautomated vehicle.

The parameter value chosen to measure driving performance affects the accuracy of the estimated fatigue level. Methods to analyze the sensitivity of these parameter values were proposed. Standard deviation of lane position (SDLP) and steering reversal rate (SRR) were considered to assess fatigue, and the sensitivity of these parameters was analyzed from the time domain and value domain. Thirty-six male drivers participated in a field test. Lane position, steering wheel angle data, and self-reported fatigue level (scored on the Karolinska sleepiness scale) were recorded. SDLP results indicate that the maximum average coefficient with fatigue level reached .11, with a unified statistical interval of 202 s when the consecutive analysis method was used; the maximum average coefficient was .12 with a unified interval of 120 s when the maximum analysis method was used. SRR results indicate that a steering angle difference of 6° was the most sensitive threshold for driver fatigue level and has an average correlation coefficient of .42, which demonstrated that SRR was more reliable than SDLP for monitoring fatigue level. With the use of the optimal parameter value, the variation results of SDLP and SRR at each fatigue level were examined, and results indicate that driving ability was impaired as fatigue level increased. The methods and results can be applied to analyses of fatigued or drowsy driving.

Current curve speed warning systems take into account mostly vehicle and road factors but not driver behavior. The systems aim at detecting sideslips of small cars on curves without consideration of rollovers for vehicles with an elevated center of gravity. In this study, a curve speed model that considers human, vehicle, and road factors is built to prevent not only sideslips but also rollover accidents for vehicles with an elevated center of gravity. In addition, a risk prediction model is presented to judge accident risk levels and determine levels of warning. Finally, the effectiveness of the presented system is validated with one skilled driver who carries out one test through a simulator under different curve scenarios. To verify the system, data from simulator tests were collected for offline checking of the system. The data were used to calculate safe speeds by using the curve speed model and to determine the levels of risk based on the risk prediction model. The results show that the system is highly compatible with the skilled driver in terms of warning accuracy and timing. Specifically, the correct alarm rate (i.e., the driver brakes and the system’s alarm goes off) of the system is 83.57% and the error alarm rate (i.e., the driver does not brake but the system’s alarm goes off) is 9.79%. Moreover, more than 80% of the time the difference between the system warning time and the operating time of the skilled driver is less than 2 s.

Some studies of driving behavior have been based on data mining to create a mechanism that relates data derived from vehicle monitoring, driver behavioral characteristics, and road safety to each other. To make the best of GPS data collected by transportation businesses and explore the potential rules of commercial vehicle driver behavioral characteristics, the parameters related to driving behavioral characteristics are extracted according to GPS data attributes based on factor analysis, and eight parameters of driving behavioral characteristics are transformed into a few aggregated variables containing clear information about driving behavior. With these variables as indicators, a cluster analysis of commercial vehicle driver behavioral characteristics in the selected case is carried out through hierarchical clustering. The results show that commercial vehicle driver behavioral characteristics can be effectively aggregated into four kinds: acceleration–deceleration, speeding-prone, acceleration, and deceleration. Of the four kinds, drivers with relatively serious acceleration–deceleration behavior are also characterized by three other relatively serious behaviors; such drivers have relatively high driving risks, so transportation businesses need to focus their supervision on those drivers. The research results have some relevance to the supervision and training of commercial vehicle drivers in China.

The parameter value chosen to measure driving performance affects the accuracy of the estimated fatigue level. Methods to analyze the sensitivity of these parameter values were proposed. Standard deviation of lane position (SDLP) and steering reversal rate (SRR) were considered to assess fatigue, and the sensitivity of these parameters was analyzed from the time domain and value domain. Thirty-six male drivers participated in a field test. Lane position, steering wheel angle data, and self-reported fatigue level (scored on the Karolinska sleepiness scale) were recorded. SDLP results indicate that the maximum average coefficient with fatigue level reached .11, with a unified statistical interval of 202 s when the consecutive analysis method was used; the maximum average coefficient was .12 with a unified interval of 120 s when the maximum analysis method was used. SRR results indicate that a steering angle difference of 6° was the most sensitive threshold for driver fatigue level and has an average correlation coefficient of .42, which demonstrated that SRR was more reliable than SDLP for monitoring fatigue level. With the use of the optimal parameter value, the variation results of SDLP and SRR at each fatigue level were examined, and results indicate that driving ability was impaired as fatigue level increased. The methods and results can be applied to analyses of fatigued or drowsy driving.

Some studies of driving behavior have been based on data mining to create a mechanism that relates data derived from vehicle monitoring, driver behavioral characteristics, and road safety to each other. To make the best of GPS data collected by transportation businesses and explore the potential rules of commercial vehicle driver behavioral characteristics, the parameters related to driving behavioral characteristics are extracted according to GPS data attributes based on factor analysis, and eight parameters of driving behavioral characteristics are transformed into a few aggregated variables containing clear information about driving behavior. With these variables as indicators, a cluster analysis of commercial vehicle driver behavioral characteristics in the selected case is carried out through hierarchical clustering. The results show that commercial vehicle driver behavioral characteristics can be effectively aggregated into four kinds: acceleration–deceleration, speeding-prone, acceleration, and deceleration. Of the four kinds, drivers with relatively serious acceleration–deceleration behavior are also characterized by three other relatively serious behaviors; such drivers have relatively high driving risks, so transportation businesses need to focus their supervision on those drivers. The research results have some relevance to the supervision and training of commercial vehicle drivers in China.

Current curve speed warning systems take into account mostly vehicle and road factors but not driver behavior. The systems aim at detecting sideslips of small cars on curves without consideration of rollovers for vehicles with an elevated center of gravity. In this study, a curve speed model that considers human, vehicle, and road factors is built to prevent not only sideslips but also rollover accidents for vehicles with an elevated center of gravity. In addition, a risk prediction model is presented to judge accident risk levels and determine levels of warning. Finally, the effectiveness of the presented system is validated with one skilled driver who carries out one test through a simulator under different curve scenarios. To verify the system, data from simulator tests were collected for offline checking of the system. The data were used to calculate safe speeds by using the curve speed model and to determine the levels of risk based on the risk prediction model. The results show that the system is highly compatible with the skilled driver in terms of warning accuracy and timing. Specifically, the correct alarm rate (i.e., the driver brakes and the system’s alarm goes off) of the system is 83.57% and the error alarm rate (i.e., the driver does not brake but the system’s alarm goes off) is 9.79%. Moreover, more than 80% of the time the difference between the system warning time and the operating time of the skilled driver is less than 2 s.

Circadian rhythms, inherent in all humans, consist of 24-h biological patterns that affect a person's fatigue level. The effect of circadian rhythms on driving performance was explored in an on-road driving study. Fifteen middle-aged professional daytime drivers were recruited to participate in the experiment. Participants were classified into three groups: (a) a morning group that started driving at 09:00, (b) a noon group that started driving at 12:00, and (c) an evening group that started driving at 21:00. Each group completed a 6-h driving task. The self-reported Karolinska sleepiness scale score was recorded every 5 min, and data on driving performance parameters, such as steering and lane positioning, were also acquired. The results indicated that both circadian rhythms and driving duration had significant effects on self-reported fatigue levels and that the fatigue level increased faster in the evening group than the morning and noon groups. The results of the circadian rhythm analysis showed that a driver was most likely to feel tired between 14:00 and 16:00 and between 02:00 and 04:00, when the ability to stay within designated lane lines (lane maintenance) was significantly impaired for drivers in all three groups. The evening group drivers were the most at risk. The steering performance did not show a significant relationship with the self-reported fatigue level. The self-reported fatigue level is the result of the interactive effect of circadian rhythms and driving duration. The standard deviation of lane position was more correlated with circadian rhythms than with the steering reversal rate.

Accurate prediction of vehicle motion status is critical for developing an advanced driver assistance system (ADAS), which can assess driving safety and detect dangerous scenarios in real time and in the near future. Although previous vehicle motion prediction models developed were mostly built on the basis of kinematic principles, driver behavior was largely ignored. Those models resulted in inaccurate trajectory predictions. To improve forecasting accuracy, the study reported here developed an improved vehicle motion model that includes consideration of both kinematic principles and real-time driver behavior. This improved vehicle motion model incorporates driver behavior into a constant acceleration (CA) model. Data on practical driver behavior, such as perception, identification, volition, and execution under traffic conditions and lane changes were collected. A quantitative approach based on a linear quadratic regulator optimal control method was used to acquire the driver's expected control input. In addition, a Kalman filter was applied to predict short-term vehicle motion, which was then used to analyze driving risks. Finally, CARSIM software was used to simulate driving scenarios. A Monte Carlo method was used to evaluate prediction accuracy and compare the results of the CA model and the improved vehicle motion model. The simulation results showed that the improved model can effectively simulate driver behavior in acceleration control by taking into full consideration the driver's volition and traffic environment. The proposed model yielded better predictions, provided an applicable way to improve the accuracy of vehicle motion prediction, and could be used to enhance the performance of ADAS.

Accurate prediction of vehicle motion status is critical for developing an advanced driver assistance system (ADAS), which can assess driving safety and detect dangerous scenarios in real time and in the near future. Although previous vehicle motion prediction models developed were mostly built on the basis of kinematic principles, driver behavior was largely ignored. Those models resulted in inaccurate trajectory predictions. To improve forecasting accuracy, the study reported here developed an improved vehicle motion model that includes consideration of both kinematic principles and real-time driver behavior. This improved vehicle motion model incorporates driver behavior into a constant acceleration (CA) model. Data on practical driver behavior, such as perception, identification, volition, and execution under traffic conditions and lane changes were collected. A quantitative approach based on a linear quadratic regulator optimal control method was used to acquire the driver's expected control input. In addition, a Kalman filter was applied to predict short-term vehicle motion, which was then used to analyze driving risks. Finally, CARSIM software was used to simulate driving scenarios. A Monte Carlo method was used to evaluate prediction accuracy and compare the results of the CA model and the improved vehicle motion model. The simulation results showed that the improved model can effectively simulate driver behavior in acceleration control by taking into full consideration the driver's volition and traffic environment. The proposed model yielded better predictions, provided an applicable way to improve the accuracy of vehicle motion prediction, and could be used to enhance the performance of ADAS.

Circadian rhythms, inherent in all humans, consist of 24-h biological patterns that affect a person's fatigue level. The effect of circadian rhythms on driving performance was explored in an on-road driving study. Fifteen middle-aged professional daytime drivers were recruited to participate in the experiment. Participants were classified into three groups: (a) a morning group that started driving at 09:00, (b) a noon group that started driving at 12:00, and (c) an evening group that started driving at 21:00. Each group completed a 6-h driving task. The self-reported Karolinska sleepiness scale score was recorded every 5 min, and data on driving performance parameters, such as steering and lane positioning, were also acquired. The results indicated that both circadian rhythms and driving duration had significant effects on self-reported fatigue levels and that the fatigue level increased faster in the evening group than the morning and noon groups. The results of the circadian rhythm analysis showed that a driver was most likely to feel tired between 14:00 and 16:00 and between 02:00 and 04:00, when the ability to stay within designated lane lines (lane maintenance) was significantly impaired for drivers in all three groups. The evening group drivers were the most at risk. The steering performance did not show a significant relationship with the self-reported fatigue level. The self-reported fatigue level is the result of the interactive effect of circadian rhythms and driving duration. The standard deviation of lane position was more correlated with circadian rhythms than with the steering reversal rate.

Planned special events attract thousands of attendees from nearby cities or suburbia by car and transit. In most cases, the majority of attendees use personal automobiles, and a high parking demand results in a short time, with a consequent parking shortage. Parking guidance information systems can solve the problem by displaying information on parking lot availability to dynamically divert vehicles. This study focused on optimizing dynamic parking guidance information for automobile drivers at special events. An original multimode traffic network was converted to a novel network by considering parking lots as dummy links; therefore the shortest path and traffic assignment could be implemented in this extended network. A bilevel programming model based on quasi-dynamic route choice and linear programming was proposed to optimize the dynamic parking guidance information. On the basis of travelers' reaction to the guidance, stochastic dynamic user optimal route choice was employed within the lower-level model. The upper-level model was a linear program aimed at minimizing network total travel time. The solutions of the bilevel programming model were based on discrete particle swarm optimization and the method of successive average algorithms. Results of a case study implemented with a hypothetical network indicated that the optimization model could reduce the system total travel time by 4%.

Planned special events attract thousands of attendees from nearby cities or suburbia by car and transit. In most cases, the majority of attendees use personal automobiles, and a high parking demand results in a short time, with a consequent parking shortage. Parking guidance information systems can solve the problem by displaying information on parking lot availability to dynamically divert vehicles. This study focused on optimizing dynamic parking guidance information for automobile drivers at special events. An original multimode traffic network was converted to a novel network by considering parking lots as dummy links; therefore the shortest path and traffic assignment could be implemented in this extended network. A bilevel programming model based on quasi-dynamic route choice and linear programming was proposed to optimize the dynamic parking guidance information. On the basis of travelers' reaction to the guidance, stochastic dynamic user optimal route choice was employed within the lower-level model. The upper-level model was a linear program aimed at minimizing network total travel time. The solutions of the bilevel programming model were based on discrete particle swarm optimization and the method of successive average algorithms. Results of a case study implemented with a hypothetical network indicated that the optimization model could reduce the system total travel time by 4%.