Bryan Higgs

A two-step segmentation algorithm for behavioral clustering of naturalistic driving styles

This research effort aims to investigate the hypothesis that drivers apply different driving styles in their daily driving tasks. A two-step algorithm is used for segmentation and clustering. First, a car-following period is broken into different duration segments that account for their temporal distribution. Second, the segments produced by the previous step are clustered based on similarity. Variations of k-means clustering and optimization techniques are used in this process. The segments centroids, used for clustering, are 8-dimensional and are produced by taking the average of the data points in each segment based on longitudinal acceleration, lateral acceleration, gyro (yaw rate), vehicle speed, lane offset, gamma (yaw angle), range, and range rate. The results of this methodology are continuous segments of car-following behavior as well as clusters of segments that show similar data and thus similar behaviors. The sample used in this paper included three different truck drivers that are representative of a high-risk driver, a medium-risk driver, and a low-risk driver. . In summary, the results revealed behavior that changed within a car-following period, between car-following periods, and between drivers. Each driver showed a unique distribution of behavior, but some of the behaviors existed in more than one driver but at different frequencies.

Integrated Real-Time Data Collection and Safety Improvement System at Signalized Intersections

Vehicles approaching a traffic signal at the onset of yellow indication will have to either stop or go. Vehicles that are far from the stopbar tend to stop, while vehicles that are close to the stopbar tend to go. Vehicles in the middle are said to be in the dilemma zone. Right-angle crashes can occur as a result of dilemma zone when drivers decide to continue even though there is not enough clearance interval for them to clear the intersection. On the other hand, rear-end crashes might occur if a vehicle stops and the following vehicle proceeds. In this paper, we present a second-generation real-time intersection safety evaluation system that was developed as a part of a Virginia Department of Transportation Project. Analysis tools utilizing the data collection and evaluation system are presented in this paper as well. These tools are currently used to develop new dilemma zone analysis concepts and evaluation metrics.

Identification of warning signs in truck driving behavior before safety-critical events

This research effort aims to shed light upon the behaviors that drivers show during safety-critical events. The 100-Car Naturalistic Driving Study conducted by the Virginia Tech Transportation Institute collected useful data about such behaviors. By instrumenting automobiles for the study and allowing them to be used during normal daily routines, the data collected included normal driving and safety-critical events. This allowed the two data sets to be compared in order to ascertain differences. A discriminant analysis was used for this task, which resulted in interesting results when analyzing the data immediately before safety-critical events for two drivers. The discriminant analysis resulted in a way to distinguish safety-critical event behavior as the discriminant scores of the data immediately before a safety-critical event show a deviation from normal car-following behavior.

Agent-based evaluation of driver heterogeneous behavior during safety-critical events

Heterogeneous driver behavior during safety-critical events is more complicated than normal driving situations and is difficult to capture by statistical models. This paper applies an agent-based reinforcement learning method to represent heterogeneous driving behavior for different drivers during safety-critical events. The naturalistic driving data of different drivers during safety-critical events are used in agent training. As an output of the Neuro-Fuzzy Actor Critic Reinforcement Learning (NFACRL) training technique, behavior rules are embedded in different agents to represent heterogeneous actions between drivers. The results show that the NFACRL is able to simulate naturalistic driver behavior and present heterogeneity.

Experimental Design for a Psychophysiological Driving Simulator Study

This research effort describes a methodology to investigate the effects of personality and emotion on driver behavior. The methodology detailed in this paper has four main components: development of psychological personality surveys, development of psychological emotion surveys, physiological data collection, and development of simulator scenarios. There were three simulator scenarios used in this study: a base scenario, an anxiety scenario, and an anger scenario. The base scenario allowed the participants to interact without any pressure being applied to them. The anxiety scenario applied pressure to the participants through a time limit and police vehicles patrolling the roadway. The anger scenario applied pressure to the participants through aggressive actions by the surrounding traffic. The results show that there can be drastic differences in driver behavior in different emotional states. The tendencies are that drivers become more aggressive and show smaller headways when they are angered. The results also show that drivers have different emotional responses to the same stimuli.

Combined car-following and unsafe event trajectory simulation using agent based modeling techniques

This paper presents a research effort aimed at modeling normal and safety-critical driving behavior in traffic under naturalistic driving data using agent based modeling techniques. Neuro-fuzzy reinforcement learning was used to train the agents. The developed agents were implemented in the VISSIM simulation platform and were evaluated by comparing the behavior of vehicles with and without agent behavior activation. The results showed very close resemblance of the behavior of agents to driver data.

Determination and optimization of reinforcement learning parameters for driver actions in traffic

An agent-based, artificial intelligence technique known as reinforcement learning has been used to capture vehicle behavior and simulate drivers actions in traffic, especially during emergency situations. This paper discusses the training parameters and their influence on agent simulation performance. A type of agent training technique called Neuro-Fuzzy Actor Critic Reinforcement Learning (NFACRL) is used to test the training parameters with an objective of improving simulation performance. A systematic parameter determination and optimization methodology is provided.

A revised reinforcement learning algorithm to model complicated vehicle continuous actions in traffic

An agent-based multi-layer reinforcement learning (RL) framework for naturalistic driving behavior simulation in traffic is introduced. Each agent is a replication of an individual driver. Each agent is implemented by applying artificial intelligence concepts, including: fuzzy logic, neural networks, and reinforcement learning algorithms. A revised Neuro-Fuzzy Actor Critic Reinforcement Learning (NFACRL) is proposed to simulate vehicle actions during safety-critical events when the traffic state is complicated. The revised NFACRL algorithm can handle state dimension problems and continuous vehicle actions.

Development of an Emotional Car-following Model

This research effort proposes a new car-following model that includes the effects of emotion on driver behavior. The methodology used to investigate the effects of emotion on driver behavior has three main components: a psychological personality survey, a psychological emotion survey, and simulator scenarios. There were three simulator scenarios used in this study: a base scenario, an anxiety scenario, and an anger scenario. The design of the study was such that the scenarios induced certain emotions in the participants and those emotions were measured using psychological emotion surveys. The base scenario allowed the participants to interact without any pressure being applied to them. The anxiety scenario applied pressure to the participants through a time limit and police vehicles patrolling the roadway. The anger scenario applied pressure to the participants through aggressive actions by the surrounding traffic. The findings showed that emotions caused drivers to change the distribution of their driving behaviors. The new car-following model uses a markovian process to account for the different distributions of behavior for each emotional state.

Identification and Classification of State-action Clusters of Car-following Behavior

This research effort aims to identify and classify state-action clusters of driver behavior. The methodology first segments and clusters car-following periods into clusters that identify a specific combination of state variables (speed, lane offset, yaw angle, range and range rate) and action variables (longitudinal acceleration, lateral acceleration, and yaw rate). The state-action clusters are then analyzed using discriminant analysis to reveal the clusters that can be identified using: (1) only state variables, (2) only action variables, and (3) both state and action variables. The sample used in this paper included ten different drivers with over 100 car-following periods each, totaling over 1500 car-following periods. In summary, the results revealed that: (1) 60% of the state-action clusters can be identified using only state variables, (2) 30% of the state-action clusters can be identified using only action variables, and 100% of the state-action clusters can be identified using both state and action variables. Also, 20% of the state-action clusters require the use of both state and action variables to be identified.