Driving Behavior Model Research Articles

Active Inference Models of AV Takeovers: Relating Model Parameters to Trust, Situation Awareness, and Fatigue.

Our objectives were to assess the efficacy of active inference models for capturing driver takeovers from automated vehicles and to evaluate the links between model parameters and self-reported cognitive fatigue, trust, and situation awareness. Control transitions between human drivers and automation pose a substantial safety and performance risk. Models of driver behavior that predict these transitions from data are a critical tool for designing safer, human-centered, systems but current models do not sufficiently account for human factors. Active inference theory is a promising approach to integrate human factors because of its grounding in cognition and translation to a quantitative modeling framework. We used data from a driving simulation to develop an active inference model of takeover performance. After validating the model's predictions, we used Bayesian regression with a spike and slab prior to assess substantial correlations between model parameters and self-reported trust, situation awareness, fatigue, and demographic factors. The model accurately captured driving takeover times. The regression results showed that increases in cognitive fatigue were associated with increased uncertainty about the need to takeover, attributable to mapping observations to environmental states. Higher situation awareness was correlated with a more precise understanding of the environment and state transitions. Higher trust was associated with increased variance in environmental conditions associated with environmental states. The results align with prior theory on trust and active inference and provide a critical connection between complex driver states and interpretable model parameters. The active inference framework can be used in the testing and validation of automated vehicle technology to calibrate design parameters to ensure safety.

A deep reinforcement learning framework for achieving super-human hazard perception in autonomous driving agents

Modeling driver behavior can significantly improve autonomous vehicles and intelligent driver assistance/training systems. Hazard perception (HP) is an essential driving skill for preventing driving accidents in potentially dangerous traffic situations. HP is not a Markov decision process, so modeling HP with reinforcement learning techniques is challenging. We introduce a framework for modeling HP scenarios using the soft actor-critic approach. Then, we use this framework to model HP in the emergence of a critical lateral hazard onto the road. The dual environments – the mirror environment for training and the target environment for usage – accelerate the learning process. The agent is employed in the target environment using an interpreter algorithm. The results show that the agent learns HP and outperforms the best human results by 2.6%. The training is implemented 40 times faster in the mirror environment. Evaluation tests show a driving experience similar to real-world scenarios rather than gaming simulations.

Modeling framework of human driving behavior based on Deep Maximum Entropy Inverse Reinforcement Learning

Driving behavior modeling is extremely crucial for designing safe, intelligent, and personalized autonomous driving systems. In this paper, a modeling framework based on Markov Decision Processes (MDPs) is introduced that emulates drivers’ decision-making processes. The framework combines the Deep Maximum Entropy Inverse Reinforcement Learning (Deep MEIRL) and a reinforcement learning algorithm-proximal strategy optimization (PPO). A neural network structure is customized for Deep MEIRL, which uses the velocity of the ego vehicle, the pedestrian position, the velocity of surrounding vehicles, the lateral distance, the surrounding vehicles’ type, and the distance to the crosswalk to recover the nonlinear reward function. The dataset of drone-based video footage is collected in Xi’an (China) to train and validate the framework. The outcomes demonstrate that Deep MEIRL-PPO outperforms traditional modeling frameworks (Maximum Entropy Inverse Reinforcement Learning (MEIRL) - PPO) in modeling and predicting human driving behavior. Specifically, in predicting human driving behavior, Deep MEIRL-PPO outperforms MEIRL-PPO by 50.71% and 43.90% on the basis of the MAE and HD, respectively. Furthermore, it is discovered that Deep MEIRL-PPO accurately learns the behavior of human drivers avoiding potential conflicts when lines of sight are occluded. This research can contribute to aiding self-driving vehicles in learning human driving behavior and avoiding unforeseen risks.

A unified risk field-based driving behavior model for car-following and lane-changing behaviors simulation

The modeling of driving behavior is pivotal for the accurate simulation of traffic scenarios and for providing human-like decision-making of autonomous driving systems. Car-following (CF) and lane-changing (LC) behaviors are continuous maneuvers within traffic flow, generally modeled separately in the literature. The coherence between these two behaviors may be ignored, leading to unrealistic behavioral simulations. Therefore, this paper establishes a risk field-based driving behavior model for two-dimensional motion, ensuring coherent modeling of CF and LC behaviors under a unified framework. First, a risk quantification method is developed to calculate the risk in two-dimensional scenarios, accounting for risk over the preview time. A cubic polynomial is applied to generate path curves that align with vehicle dynamics. Second, the enhanced behavior model primarily comprises two integral components: path and trajectory planning. These two components aim to identify the path or trajectory that maximizes the benefit while meeting the desired risk. Third, the maximum acceptable risk, representing a higher risk than the desired risk, is defined to facilitate path adjustment and avoid frequent path adjustment. Finally, the proposed model is proved through comparisons with existing models using driving data. Several cases are employed for further analysis to show the model's rationality and potential in various aspects. This study develops the previous risk field-based behavior model from one-dimensional to two-dimensional scenarios, furnishes a unified framework for elucidating driving behavior in various scenarios, and contributes to the progress of behavior modeling.

Evaluation of Traffic Efficiency and Energy-Saving Benefits of L3 Smart Vehicles under the Urban Expressway Scenario

An L3 smart vehicle (L3 SV) could behave differently from human-driven vehicles due to its intelligent configuration and decision-making logic, and may consequently exert influences on traffic flow. In order to clarify the L3 SV’s traffic impacts and accelerate L3 SV implementation, this paper conducted an evaluation on the traffic efficiency and energy consumption influences of L3 SVs based on a microscopic traffic simulation. Taking Beijing as a case, the relevant traffic economic benefits were calculated with the help of the previously proposed traffic economic benefits model. Before modeling L3 SVs, a two-dimensional general modeling architecture for SVs was proposed. According to the architecture, the driving behavior model, as well as behavior selection models of L3 SVs, was eventually determined, based on which the intelligent driving model of L3 SVs was established. Urban expressways were selected as the simulation road type, and scenario analysis was conducted on various proportions of L3 SVs and L3 connected and autonomous vehicles (L3 CAVs), as well as the input traffic flow rates. It was found that L3 SVs can significantly reduce the travel time and energy consumption and enlarge the actual road capacity. The improvement will become particularly prominent under saturated or supersaturated traffic flow and increasing the proportion of L3 SVs and L3 CAVs can also amplify the effect of traffic optimization. The related economic benefit is considerable, which is CNY 3.104 billion a year based on Beijing’s travel and traffic conditions.

Driver behavior modeling at uncontrolled intersections under Indian traffic conditions

Driver behavior modeling at uncontrolled intersections under Indian traffic conditions

Data-Driven Controller for Drivers’ Steering-Wheel Operating Behaviour in Haptic Assistive Driving System

An advanced driver-assistance system (ADAS) is critical to driver–vehicle-interaction systems. Driving behaviour modelling and control significantly improves the global performance of ADASs. A haptic assistive system assists the driver by providing a specific torque on the steering wheel according to the driving–vehicle–road profile to improve the steering control. However, the main problem is designing a compensator dealing with the high-level uncertainties in different driving scenarios with haptic driver assistance, where different personalities and diverse perceptions of drivers are considered. These differences can lead to poor driving performance if not properly accounted for. This paper focuses on designing a data-driven model-free compensator considering various driving behaviours with a haptic feedback system. A backpropagation neural network (BPNN) models driving behaviour based on real driving data (speed, acceleration, vehicle orientation, and current steering angle). Then, the genetic algorithm (GA) optimises the integral time absolute error (ITEA) function to produce the best multiple PID compensation parameters for various driving behaviours (such as speeding/braking, lane-keeping and turning), which are then utilised by the fuzzy logic to provide different driving commands. An experiment was conducted with five participants in a driving simulator. During the second experiment, seven participants drove in the simulator to evaluate the robustness of the proposed combined GA proportional-integral-derivative (PID) offline, and the fuzzy-PID controller applied online. The third experiment was conducted to validate the proposed data-driven controller. The experiment and simulation results evaluated the ITAE of the lateral displacement and yaw angle during various driving behaviours. The results validated the proposed method by significantly enhancing the driving performance.

A dynamics model for driving behavior based on coupling actuation of bounded rational cognition and diverse emotions

Precise comprehension about the impacts of drivers’ rational and perceptual characteristics on their behavioral decisions is crucial for the accurate prediction of driving behavior. In the previous research on driving behavior, drivers were regarded as homogeneous and absolutely rational individuals. To overcome this limitation, the coupling effects of bounded rational cognition and diverse emotions are considered, and a driving behavior model is proposed based on Dynamics Psychology. There are two parts in this research. In Study 1, the information entropy theory is applied to describe the vehicle cluster situation, and a method is established to quantify the cognitive uncertainty of the vehicle cluster situation for drivers in diverse emotions. In Study 2, with consideration of bounded rational and emotional cognition, the impacts of the vehicle cluster situation and drivers' features, which are their demands for the driving goals including safety, efficiency and comfort, emotional states, and cognitive characteristics, on driving behavior are uniformly expressed as the behavioral driving force, and a prediction model for driving behavior is proposed based on Dynamics Psychology. The results of validation based on the virtual driving data show the prediction accuracy of the proposed model for various driving behaviors of drivers in diverse emotions is over 80%. The results of verification based on NGSIM data suggest that the prediction accuracy of the proposed model for the natural driving behaviors is 82.07%. The research results can contribute to the study on the intrinsic mechanism of driving behavior and provide theoretical support for the development of traffic simulation, personalized active safety systems, human–machine interaction, and brain-inspired autonomous driving.

Experimental Observation and Validation of EV Model for Real Driving Behavior

Experimental Observation and Validation of EV Model for Real Driving Behavior

Instant Inverse Modeling of Stochastic Driving Behavior With Deep Reinforcement Learning

Instant Inverse Modeling of Stochastic Driving Behavior With Deep Reinforcement Learning

Driving Behavior Model for Multi-Vehicle Interaction at Uncontrolled Intersections Based on Risk Field Considering Drivers’ Visual Field Characteristics

Driving Behavior Model for Multi-Vehicle Interaction at Uncontrolled Intersections Based on Risk Field Considering Drivers’ Visual Field Characteristics

Exploring Safe Overtaking Behavior on Two‐Lane Two‐Way Road Using Multiagent Driving Simulators and Traffic Simulation

Safety and efficiency of autonomous driving behavior are a tradeoff. Behaviors that are too focused on safety can reduce road operation efficiency, while those that are too efficient can compromise passengers’ safety beyond their tolerance. Therefore, it is important to understand people’s characteristics and maintain a balance between safety and efficiency. Overtaking, which involves passing the preceding vehicle and improving road capacity, requires complex interaction as collisions with opposing vehicles must be avoided on a two‐lane, two‐way road. Overtaking to increase road capacity can induce unnecessary deceleration in oncoming vehicles, harming oncoming traffic flow. To address these concerns, a diverse dataset of natural overtaking behavior is a priority. We conduct experiments using a network connection between two multiagent driving simulators to collect a human behavior‐based overtaking dataset and develop driving behavior models engaged in overtaking situations using the Extra Trees model. The behavior models are embedded in microsimulation to generate human behavior‐based datasets under different conditions using a dynamic link library and component object model interfaces. To understand the interaction in an overtaking scenario by the generated datasets, we used a K‐means clustering technique to analyze the different reaction behaviors between the oncoming and overtaking vehicles. The threshold for achieving a balanced combination of safety and efficiency is established using XGboost. Finally, safe overtaking behavior is analyzed using a combination of the classified driving styles and thresholds. The results show that the overtaking vehicle can safely start overtaking without endangering oncoming vehicles when both speed and distance conditions are met simultaneously; the speed is lower than 44.29 km/h and it is 407 m away from oncoming vehicles.

Narrow passage interactions: A UK-based exploratory survey study to identify factors affecting driver decision-making

Narrow passage interactions have received increased attention from academics seeking to create behavioural models of the interaction and those looking to define how autonomous vehicles (AVs) should interact with their human counterparts in a composite road system. Despite this increased attention, many factors remain unexplored in the narrow passage literature, with the literature also encompassing few driving culture contexts. To this end, this study employs an explorative survey to identify additional factors that affect driver decision-making during narrow passage interactions, as well as driver perceptions of different communications in a UK context. The study’s 243 participants were presented with a range of different narrow passage scenarios and asked to indicate how likely they were to give way/yield to a vehicle approaching the narrow passage from the opposite direction. In addition, they also completed the Multidimensional Driving Style Inventory to identify their driving styles and asked to identify which signals they look for from their interaction partner during narrow passage interactions, as well as the meaning of those signals. The results of the study show that situational characteristics such as the vehicle type being interacted with, being in a rush and being followed by vehicles alter the likelihood of drivers giving way at narrow passages, whilst a person’s driving style can also indicate how likely someone is to give way to another vehicle. These results highlight the factors that are considered by drivers, increasing our understanding of the factors that need to be incorporated in driver behaviour models and in AV development.

Modelling Driver Behaviour at Urban Signalised Intersections Using Logistic Regression and Machine Learning

This study investigated several factors that may influence driver actions throughout the yellow interval at urban signalised intersections. The selected samples include 2,168 observations. Almost 33% of drivers stopped ahead of the stop line, 60% passed the intersection through the yellow interval, and 7% passed after the yellow interval was complete (red light running, RLR violations). Binary logistic regression models showed that the chance of passing went up as vehicle speed went up and down as the gap between the vehicle and the traffic light and green interval went up. The movement type and vehicle position influenced the passing probability, but the vehicle type did not. Moreover, multinomial logistic regression models showed that the legal passing probability declined with the growth in the green time and vehicle distance to the traffic signal. It also increased with the growth in the speed of approaching vehicles. Also, movement type directly affected the chance of legally passing, but vehicle position and type did not. Furthermore, the driver’s performance during the yellow phase was studied using the k-nearest neighbours algorithm (KNN), support vector machines (SVM), random forest (RF) and AdaBoost machine learning techniques. The driver’s action run prediction was the most accurate, and the run-on-red camera was the least accurate.

A Personalized Motion Planning Method with Driver Characteristics in Longitudinal and Lateral Directions

Humanlike driving is significant in improving the safety and comfort of automated vehicles. This paper proposes a personalized motion planning method with driver characteristics in longitudinal and lateral directions for highway automated driving. The motion planning is decoupled into path optimization and speed optimization under the framework of the Baidu Apollo EM motion planner. For modeling driver behavior in the longitudinal direction, a car-following model is developed and integrated into the speed optimizer based on a weight ratio hypothesis model of the objective functional, whose parameters are obtained by Bayesian optimization and leave-one-out cross validation using the driving data. For modeling driver behavior in the lateral direction, a Bayesian network (BN), which maps the physical states of the ego vehicle and surrounding vehicles and the lateral intentions of the surrounding vehicles to the driver’s lateral intentions, is built in an efficient and lightweight way using driving data. Further, a personalized reference trajectory decider is developed based on the BN, considering traffic regulations, the driver’s preference, and the costs of the trajectories. According to the actual traffic scenarios in the driving data, a simulation is constructed, and the results validate the human likeness of the proposed motion planning method.

Application of Game Refinement Theory to Automated Driving

In recent years, assisted driving and self-driving have captured the imagination of manufacturers, designers, technology providers, and the general public with the expectation of a sustainable, safer, and intelligent mobility in the near future. Self-driving or assisted driving vehicles are complex systems that integrate environmental perception, intelligent planning and decision-making, tracking, and control. With the increasing intelligence of vehicles, personalized design is an inevitable trend. A design that is in line with the driver’s personality can bring a better driving experience to the driver. Thus, classifying driving types while driving in a self-driving environment may play an important role in the construction of trajectory planning algorithms. This paper uses the motion-in-mind model from game refinement theory to model driver behavior. Further, a classification of the model parameters into three categories helped in distinguishing cautious, aggressive, and average drivers. The results showed that the self-driving environment can be successfully modeled as a game and adaptation to match the riders’ driving skills may improve satisfaction.

Fuzzy modelling and inference for physics-aware road vehicle driver behaviour model calibration

Traffic simulation is a crucial tool for testing and evaluating hypotheses in both the design of autonomous driving systems and the optimization of urban transportation, where modelling and calibrating driver behaviour is key. In this article, we introduce a novel approach to calibrating driver models by combining fuzzy logic and inference to produce a plausible parametrization that, simultaneously, reproduces drivers’ behaviour peculiarities and road traffic characteristics. The calibration system uses solely vehicle trajectories that describe cars’ movements through the road network, with a resolution of time and space enough to allow for determining the lane of the road used, speed, and acceleration at every second of the survey period. This data is processed by a lightweight modelling and inference system capable of interpreting the input data through the lens of domain knowledge and physics laws in order to learn plausible driver model parameters using fuzzy computation. The outcome is a set of parameters for mathematical models which emulate the social and physical features of the behaviour of a driver-vehicle unit. Albeit this study limits itself to determining parameters for car-following and lane-change models, the systematic approach for calibration can be seen as a framework where also other driver model components can be calibrated (e.g. gap-acceptance). We evaluate our system on both synthetic and real data while comparing it with other approaches for calibration. We summarize our findings using a thorough analysis where we point out the most relevant aspects to consider when calibrating driver models.

A risk‐based driver behaviour model

AbstractCurrent driver behaviour models (DBMs) are primarily designed for the general driver population under specific scenarios, such as car following or lane changing. Hence DBMs capturing individual behaviour under various scenarios are lacking. This paper presents a novel method to quantify individual perceived driving risk in the longitudinal and lateral directions using risk thresholds capturing the time headway and time to line crossing. These are integrated in a risk‐based DBM formulated under a model predictive control (MPC) framework taking into account vehicle dynamics. The DBM assumes drivers to operate as predictive controllers jointly optimising multiple criteria, including driving risk, discomfort, and travel inefficiency. Simulation results in car following and passing a slower vehicle demonstrate that the DBM predicts plausible behaviour under representative driving scenarios, and that the risk thresholds are able to reflect individual driving behaviour. Furthermore, the proposed DBM is verified using empirical driving data collected from a driving simulator, and the results show it is able to accurately generate vehicle longitudinal and lateral control matching individual human drivers. Overall, this model can capture individual risk perception behaviour and can be applied to the design and assessment of intelligent vehicle systems.

Simulation-Based Testing and Performance Evaluation of Vehicle Safety Functions

<div>The progressive development toward highly automated driving poses major challenges for the release and validation process in the automotive industry, because the immense number of test kilometers that have to be covered with the vehicle cannot be tackled to any extent with established test methods, which are highly focused on the real vehicle. For this reason, new methodologies are required. Simulation-based testing and, in particular, virtual driving tests will play an important role in this context. A basic prerequisite for achieving a significant reduction in the test effort with the real vehicle through these simulations are realistic test scenarios. For this reason, this article presents a novel approach for generating relevant traffic situations based on a traffic flow simulation in SUMO and a vehicle dynamics simulation in CarMaker. The procedure is shown schematically for an emergency braking function. A driving function under test faces the major challenges when the other road users commit driving errors. Therefore, the driving behavior models in this traffic flow simulation are modified in such a way that critical scenarios can arise because of these driving errors. In order to be able to make a statement about the correct behavior of the driving function under test in these traffic situations, objective criteria are necessary to evaluate the triggering behavior and the handling of the traffic situations. Based on the performance evaluation of the driving function under test, characteristic test scenarios are then identified that evenly cover the test space. The comparison of the deviations in covering this test space with full and the reduced dataset is small except in areas where there are no scenarios in both datasets. Finally, these selected scenarios are used to perform an application of the driving function under test. The procedure is exemplified for the triggering time and the maximum deceleration of an emergency braking function. When comparing the distributions, it is shown that the performance in both datasets improves in the same way when parameters are optimized. For example, the mean performance of the driving function increases by more than 0.3 in each case when optimizing the triggering time. Thus, it is no longer necessary to use all scenarios for parameterization in virtual driving tests.</div>

Is trust a belief, observation, or state: Results from an active inference analysis of driver-automation transitions of control

Transitions of control are an important safety concern for human-automation teams and automated vehicle safety. While trust and situation awareness have been observed to influence transitions of control in automated vehicles, there are few objective measurements, making these concepts difficult to operationalize in increasingly automated decision systems. In this study, we take a step towards quantifying trust by mapping latent driver beliefs extracted from an active inference-factor analysis model of driver behavior and cognitive dynamics to subjective responses to trust questionnaires. Our results show that subjective trust is primarily correlated with model parameters affecting perceptual evidence accumulation rate, and the same parameters are significantly correlated with driver age.