Indrajit Chatterjee

Outline for a causal model of traffic conflicts and crashes

Road crashes tend to be infrequent but with nontrivial consequences, leading to a long-running interest in identifying surrogate events, such as traffic conflicts, that can support a timely recognition and correction of safety deficiencies. Although a variety of possible surrogates have been suggested, questions remain regarding how crashes and surrogates are related. Using recent developments in causal analysis we propose a simple model which represents crashes and conflicts as resulting from interactions between initiating conditions and evasive actions, and then use this model to identify relationships between these types of events. Our first set of results expresses the probability of a crash as a mixture of probabilities over different sets of initiating conditions, where the mixing probabilities are governed by the evasive action. Our second set of results considers situations where sampling is restricted to non-crash events, and gives conditions where these truncated probabilities can serve as proxies for crash probabilities. We then illustrate how trajectory-based reconstruction can be used to classify initiating events with respect to severity, and to estimate a proxy for the crash probability from a set of incompletely observed non-crash events.

Analysis of Rear-End Events on Congested Freeways by Using Video-Recorded Shock Waves

Rear-end crashes on congested freeways are among the most frequently occurring capacity-reducing incidents. Empirical evidence suggests that freeway rear-end crashes tend to occur when stopping shock waves form on freeways. To understand the occurrence of such incidents, this study determined conditions by which a stopping wave results in a rear-end crash. Reviewing the existing literature on rear-end crash mechanisms, the study first established a sufficient condition for a shock wave to produce a rear-end crash and then verified it by using video recordings of 41 shock waves, including five resulting in rear-end crashes, on Interstate I-94 in Minneapolis, Minnesota. For each shock wave, key kinematic features of the involved drivers, including initial speeds, following headway, average decelerations, and reaction times, were estimated. The proposed crash condition successfully distinguished between successful brake-to-stop events and rear-end crashes. Detailed analysis of two rear-end crashes illustrated how drivers with reaction times longer than their following headways played a key role in determining the crash outcome of a shock wave. These findings suggest that interventions focusing on the relationship between reaction time and following headway could reduce the frequency of rear-end crashes.

A statistical process control approach using cumulative sum control chart analysis for traffic data quality verification and sensor calibration for weigh-in-motion systems

ABSTRACT Weigh-in-motion systems have been widely used by state agencies to collect the traffic data on major state roadways and bridges to support traffic load forecasting, pavement design and analysis, infrastructure investment decision making, and transportation planning. However, the weigh-in-motion system itself poses difficulties in obtaining accurate data due to sensor characteristics that can be sensitive to vehicle speed, weather conditions, and changes in surrounding pavement conditions. This study focuses on developing a systematic methodology to detect weigh-in-motion sensor bias and enhance current practices for weigh-in-motion calibration. A mixture modeling technique using an expectation maximization algorithm was developed to divide the vehicle class 9 gross vehicle weight into three normally distributed components: unloaded, partially loaded, and fully loaded trucks. Then the well-known statistical process control technique cumulative sum control chart analysis was applied to expectation maximization estimates of daily mean gross vehicle weight for fully loaded trucks to identify and estimate shifts in the weigh-in-motion sensor. Special attention was given to the presence of autocorrelation in the data by fitting an autoregressive time-series model and then performing cumulative sum control chart analysis on the fitted residuals. Results from the analysis suggest that the proposed methodology is able to estimate a shift in the weigh-in-motion sensor accurately and also indicate the time point when the system went out of calibration. This methodology can be effectively implemented by state agencies, resulting in more accurate and reliable weigh-in-motion data.

Use of Naturalistic Driving Data to Characterize Driver Behavior in Freeway Shockwaves

Recent years have witnessed significant efforts in developing and evaluating vehicle-based passive and active safety systems to reduce traffic accidents. In addition, there is growing interest in the use of microscopic simulation models for evaluating operational strategies. Both activities require quantitative characterization of driver behavior in real-world situations. Historically, such characterizations have been difficult to obtain, but the data available from large-scale naturalistic driving studies (NDS) have the potential to change this situation. However, identifying relevant events from an NDS database and reducing the NDS data to estimate relevant features of the events are still something of a challenge. This study used freeway brake-to-stop events on congested freeways as examples to describe methods for identifying relevant events. It then estimated event features, such as initial speeds for leading and following vehicles, reaction times for leading and following drivers, and changes in the drivers’ braking rates. A suitably representative sample of such estimates could be used to support evaluation of vehicle-based safety countermeasures or provide inputs to traffic simulation models.

Evolutionary Game Theoretic Approach to Rear-End Events on Congested Freeway

Rear-end crashes on freeways contribute significantly to nonrecurring congestion. Reducing these events would significantly improve freeway capacity, particularly during peak hours. Although promising countermeasures, such as variable speed limits, changeable message signs, and vehicle-based improvements, are under consideration, currently there is a shortage of demonstrably proven countermeasures targeted at freeway rear-end crashes. Liability rules, in which the direct cost associated with a crash is divided between the drivers, their insurance companies, or both, are a primary mechanism for influencing the occurrence of freeway rear-end crashes. An exploratory effort uses concepts from evolutionary game theory to predict the effects of liability rules on rear-end crashes. In a typical two-vehicle car-following scenario, driving behavior can be associated with a utility that each driver expects to achieve depending on his or her and the opponents actions. Such interactions between leader and follower are modeled as the outcome of an evolutionary process in which drivers with different driving behaviors are randomly and repeatedly matched against each other to play a two-player game. The outcome of these games determines the fraction of drivers pursuing a particular driving strategy for the next phase of the game. The stable long-run distribution of driving strategies is then used to predict the proportion of drivers who are more likely to be involved in a rear-end accident. It turns out that when direct crash costs are allocated evenly to the involved drivers, a population in which all drivers act to avoid crashes is not evolutionarily stable.