Intelligent Driver Research Articles

Bifurcation control based on improved intelligent driver model considering stability and minimum gasoline consumption

Advanced driver assistance system (ADAS) plays an important role in the transition period before the full maturity of self-driving technology. In this paper, an intelligent driver model with multiple time delays (IDM-MTD) is established to describe the dynamic characteristics of vehicles equipped with ADAS. Moreover, a control term that considers the velocity difference between the current and historical moment is selected for assistance driving. A novel framework is proposed to determine control parameters in controlled IDM-MTD. The definite integral stability method (DISM) is integrated into the framework to constrain the range of parameters that make the system stable. In the control strategy design process, this paper places significant emphasis on the implementation of the DISM modification to effectively address redundant iteration steps by exploiting the characteristic of bifurcation. Through numerical simulation, a two-dimensional optimal parameter combination is found, which not only stabilizes traffic flow but also produces minimal gasoline consumption.

Evaluating the Effect of Road Surface Potholes Using a Microscopic Traffic Model

Road surface wear leads to the formation of cracks and holes known as potholes. Potholes disrupt the smooth flow of traffic and can lead to accidents. The Intelligent Driver (ID) model is commonly employed but it assumes uniform traffic behavior for all conditions. This oversimplified approach is unrealistic as it does not consider the impact of real-world factors such as potholes on traffic patterns. This paper proposes a microscopic traffic model to address the impact of these road surface irregularities on traffic. The effect of small, medium, and large conical potholes is investigated using fundamental diagrams for traffic flow and velocity. The results obtained indicate that the proposed model outperforms the ID model as it can more accurately characterize how potholes and driver sensitivity affect vehicle behavior.

A Survey on Time-Sensitive Networking Standards and Applications for Intelligent Driving

Stimulated by the increase in user demands and the development of intelligent driving, the automotive industry is pursuing high-bandwidth techniques, low-cost network deployment and deterministic data transmission. Time-sensitive networking (TSN) based on Ethernet provides a possible solution to these targets, which is arousing extensive attention from both academia and industry. We review TSN-related academic research papers published by major academic publishers and analyze research trends in TSN. This paper provides an up-to-date comprehensive survey of TSN-related standards, from the perspective of the physical layer, data link layer, network layer and protocol test. Then we classify intelligent driving products with TSN characteristics. With the consideration of more of the latest specified TSN protocols, we further analyze the minimum complete set of specifications and give the corresponding demo setup for the realization of TSN on automobiles. Open issues to be solved and trends of TSN are identified and analyzed, followed by possible solutions. Therefore, this paper can be an investigating basis and reference of TSN, especially for the TSN on automotive applications.

A Microscopic Traffic Model Considering Driver Reaction and Sensitivity

A new microscopic traffic model is proposed that characterizes driver response according to reaction and sensitivity. Driver response in the intelligent driver (ID) model is based on a fixed acceleration exponent and so does not follow traffic physics. This inadequate characterization results in unrealistic traffic behavior. With the proposed model, drivers can be aggressive, sluggish, or typical. It is shown to be string stable, and for appropriate distance headway and velocity (speed), the traffic flow is smooth. Furthermore, the proposed model has better stability than the ID model because it is based on driver reaction and sensitivity, while the ID model is based on a fixed exponent. The ID and proposed models are evaluated on a circular road of length 1200 m with a platoon of 21 vehicles for 150 s. The results obtained show that the proposed model characterizes traffic more realistically than the ID model.

Fuel Consumption Evaluation of Connected Automated Vehicles Under Rear-End Collisions

Connected automated vehicles (CAV) can increase traffic efficiency, which is considered a critical factor in saving energy and reducing emissions in traffic congestion. In this paper, systematic traffic simulations are conducted for three car-following modes, including intelligent driver model (IDM), adaptive cruise control (ACC), and cooperative ACC (CACC), in congestions caused by rear-end collisions. From the perspectives of lane density, vehicle trajectory and vehicle speed, the fuel consumption of vehicles under the three car-following modes are compared and analysed, respectively. Based on the vehicle driving and accident environment parameters, an XGBoost algorithm-based fuel consumption prediction framework is proposed for traffic congestions caused by rear-end collisions. The results show that compared with IDM and ACC modes, the vehicles in CACC car-following mode have the ideal performance in terms of total fuel consumption; besides, the traffic flow in CACC mode is more stable, and the speed fluctuation is relatively tiny in different accident impact regions, which meets the driving desires of drivers.

Safe, Efficient, and Comfortable Reinforcement-Learning-Based Car-Following for AVs with an Analytic Safety Guarantee and Dynamic Target Speed

Over the last decade, there has been rising interest in automated driving systems and adaptive cruise control (ACC). Controllers based on reinforcement learning (RL) are particularly promising for autonomous driving, being able to optimize a combination of criteria such as efficiency, stability, and comfort. However, RL-based controllers typically offer no safety guarantees. In this paper, we propose SECRM (the Safe, Efficient, and Comfortable RL-based car-following Model) for autonomous car-following that balances traffic efficiency maximization and jerk minimization, subject to a hard analytic safety constraint on acceleration. The acceleration constraint is derived from the criterion that the follower vehicle must have sufficient headway to be able to avoid a crash if the leader vehicle brakes suddenly. We critique safety criteria based on the time-to-collision (TTC) threshold (commonly used for RL controllers), and confirm in simulator experiments that a representative previous TTC-threshold-based RL autonomous-vehicle controller may crash (in both training and testing). In contrast, we verify that our controller SECRM is safe, in training scenarios with a wide range of leader behaviors, and in both regular-driving and emergency-braking test scenarios. We find that SECRM compares favorably in efficiency, comfort, and speed-following to both classical (non-learned) car-following controllers (intelligent driver model, Shladover, Gipps) and a representative RL-based car-following controller.

Evaluation of car-following model for inland vessel-following behavior

Due to the unique role of inland waterway transport in reducing carbon emissions, the demand for inland vessel transportation has greatly increased recent years in China. More and more vessels in the inland waterway have gradually formed a non-free following queue, rather than the free traffic flow of the past. To provide a theoretical basis for organizing the vessel traffic under a crowded state, it is necessary to understand accurately the microscopic characteristics of inland waterway vessel traffic flow, such as vessel-following behavior. This research carried out three vessel-following experiments to obtain valuable microscopic data of vessel-following behavior in different inland restricted waterways in Yangtze Delta, China. Considering the uniformity of traffic flow, six typical car-following models were tried to verify their applicability to inland vessel-following behavior. The parameters of the car-following models were recalibrated based on the experimental data. The simulation results of different models were evaluated, and the improved General Motor (GM) model was proposed to avoid vessel collision during the simulation process in GM model. The results show that the improved GM model, Newell model, Intelligent Driver model (IDM) and Longitudinal Control model (LCM) were confirmed to be suitable to describe the vessel-following behavior in the inland restricted waterway. These models may help professionals understand deeply the inland vessel traffic characteristics.

Multiple Shape Manipulation of Azobenzene-Containing Polyimide by Combining Shape Memory Effect, Photofixity, and Photodeformation.

The integration of different shape manipulation could greatly expand the versatility and functionality of smart materials, for which the achievement of synergism of different shape control is crucial. Here, we seek to create one kind of polyimide with integrated multiple shape manipulations by constructing the chemical network bearing azobenzene as a side chain. Trifunctional cross-linkers serving as net points of the chemical network render polyimide thermal-induced shape memory effects, which enables shape transformation. Azobenzene as a photoresponsive group is employed to achieve the photofixity and reversible photodeformability. Such photosensitive behaviors are independent of molecular prealignment and remain available after thermally shaping and fixing. As a result, these noninterfering performances induced by heat and light allow us to arbitrarily combine them to meet different needs. By integrating different shape manipulations, various shape changes and functional execution are conveniently achieved. The combination of the shape memory effect with photofixity enables the setting of diverse shapes, while the merging of it with reversible deformation facilitates the construction of actuators capable of executing functions. This study provides a new approach for the preparation of multifunctional actuators and has potential applications in the field of intelligent drivers.

High-Dynamic-Range Tone Mapping in Intelligent Automotive Systems

Intelligent driver assistance systems are becoming increasingly popular in modern passenger vehicles. A crucial component of intelligent vehicles is the ability to detect vulnerable road users (VRUs) for an early and safe response. However, standard imaging sensors perform poorly in conditions of strong illumination contrast, such as approaching a tunnel or at night, due to their dynamic range limitations. In this paper, we focus on the use of high-dynamic-range (HDR) imaging sensors in vehicle perception systems and the subsequent need for tone mapping of the acquired data into a standard 8-bit representation. To our knowledge, no previous studies have evaluated the impact of tone mapping on object detection performance. We investigate the potential for optimizing HDR tone mapping to achieve a natural image appearance while facilitating object detection of state-of-the-art detectors designed for standard dynamic range (SDR) images. Our proposed approach relies on a lightweight convolutional neural network (CNN) that tone maps HDR video frames into a standard 8-bit representation. We introduce a novel training approach called detection-informed tone mapping (DI-TM) and evaluate its performance with respect to its effectiveness and robustness in various scene conditions, as well as its performance relative to an existing state-of-the-art tone mapping method. The results show that the proposed DI-TM method achieves the best results in terms of detection performance metrics in challenging dynamic range conditions, while both methods perform well in typical, non-challenging conditions. In challenging conditions, our method improves the detection F2 score by 13%. Compared to SDR images, the increase in F2 score is 49%.

A Microscopic Traffic Model Considering Time Headway and Distance Headway

A microscopic traffic model is presented which employs differences in velocity to characterize driver behavior. The Intelligent Driver (ID) model is based on an acceleration constant which cannot capture different traffic conditions. Further, it is not based on traffic physics and so can produce inaccurate results. The proposed model is an improved ID model and both are evaluated on a 2000 m circular road. The results obtained show that the proposed model can appropriately characterize traffic flow and density. Further, the variations in flow and velocity are smoother than with the ID model. This is because the proposed model is based on actual traffic parameters rather than an unrealistic traffic exponent.

Machine learning based driver monitoring system: A case study for the Kayoola EVS

With the ever-growing traffic density, the number of road accidents has continued to increase. Finding solutions to reduce road accidents and improve traffic safety has become a top priority for Kiira Motors Corporation, a Ugandan state-owned automotive company. The company seeks to develop intelligent driver assistance systems for its market entry product, the Kayoola EVS bus. A machine learning-based driver monitoring system that would monitor driver drowsiness and send out an alarm in case drowsiness is detected has been developed in an attempt to reduce drowsiness-related accidents. The system consists of a camera positioned in such a way as to keep track of the driver's face. The camera is interfaced with a Raspberry Pi minicomputer which carries out the computations and analysis and when drowsiness is detected, an alarm is triggered. Dangerous driver behavior including distraction and fatigue has long been recognized as the main contributing factor in traffic accidents. This paper therefore presents the development of a driver monitoring system for the Kayoola Electric City Bus to address the increasing occurrences of road accidents. The machine learning-based driver monitoring system is designed to be non-intrusive with continuous real-time operation.

Identifying suitable car-following models to simulate automated vehicles on highways

This study addresses car-following models that are currently used for simulating AV and CAV. Diverse car-following models, Intelligent Driver Model (IDM), Improved IDM (IIDM), IIDM with Constant-Acceleration Heuristic (CAH), and MIcroscopic model for Simulation of Intelligent Cruise control (MIXIC) are examined with the state-of-the-art vehicle trajectory data, Highway Drone dataset (HighD), and genetic algorithm. There is no commercial level 5 AV or CAV as of 2022; therefore, the authors generate hypothetical AV trajectories based on the actual vehicle trajectories and the assumption of an ideal AV. Based on the analysis, the calibrated IIDM with CAH shows the most fit on AV behavior.

Improved deep reinforcement learning for car-following decision-making

Accuracy improvement of Car-following (CF) model has attracted much attention in recent years. Although a few studies incorporate deep reinforcement learning (DRL) to describe CF behaviors, proper design of reward function is still an intractable problem. This study improves the deep deterministic policy gradient (DDPG) car-following model with stacked denoising autoencoders (SDAE), and proposes a data-driven reward representation function, which quantifies the implicit interaction between ego vehicle and preceding vehicle in car-following process. The experimental results demonstrate that DDPG-SDAE model has superior ability of imitating driving behavior: (1) validating effectiveness of the reward representation method with low deviation of trajectory; (2) demonstrating generalization ability on two different trajectory datasets (HighD and SPMD); (3) adapting to three traffic scenarios clustered by a dynamic time warping distance based k-medoids method. Compared with Recurrent Neural Networks (RNN) and intelligent driver model (IDM), DDPG-SDAE model shows better performance on the deviation of speed and relative distance. This study demonstrates superiority of a novel reward extraction method fusing SDAE into DDPG algorithm and provides inspiration for developing driving decision-making model.

Bayesian Networks for the Driver Overtaking Assistance System on Two-lane Roads

Unsuccessful overtaking maneuvers on two-lane rural roads are one of the major causes of road accidents in the 21st century. The complexity of this maneuver merits the adoption of a thorough method for developing a proposed assistance system to prevent accidents and consequently reduce the high number of fatalities and the associated economic costs. This study aims to introduce an intelligent Driver Overtaking Assistance System (DOAS) to assist drivers in performing overtaking maneuvers safely. The study also will introduce a method to assess the impact of all the influential variables related to the driver, vehicle, traffic, road, and the surrounding environment. In momentary driving situations, the DOAS uses the communicated information via Hello beacon messages (HBM) and a set of input sensors to measure the possibility of overtaking the preceding vehicle(s) proactively by considering whether the distance gap to the oncoming vehicle is sufficient for overtaking. Besides, the proposed system is a vehicle-based safety system based on the collection of contextual information from the driving vicinity to acquire all relevant information regarding the ambient driving environment and the vehicles involved in the overtaking. To do this, DOAS uses a Bayesian Network (BN) to model overtaking maneuvers. The work presented shows high accuracy and promising results in aiding safe overtaking, with significant improvements to overtaking maneuvers on two-lane rural roads.

Calibration Measurements and Computational Models of Sensors Used in Autonomous Vehicles

Unsuccessful overtaking maneuvers on two-lane rural roads are one of the major causes of road accidents in the 21st century. The complexity of this maneuver merits the adoption of a thorough method for developing a proposed assistance system to prevent accidents and consequently reduce the high number of fatalities and the associated economic costs. This study aims to introduce an intelligent Driver Overtaking Assistance System (DOAS) to assist drivers in performing overtaking maneuvers safely. The study also will introduce a method to assess the impact of all the influential variables related to the driver, vehicle, traffic, road, and the surrounding environment. In momentary driving situations, the DOAS uses the communicated information via Hello beacon messages (HBM) and a set of input sensors to measure the possibility of overtaking the preceding vehicle(s) proactively by considering whether the distance gap to the oncoming vehicle is sufficient for overtaking. Besides, the proposed system is a vehicle-based safety system based on the collection of contextual information from the driving vicinity to acquire all relevant information regarding the ambient driving environment and the vehicles involved in the overtaking. To do this, DOAS uses a Bayesian Network (BN) to model overtaking maneuvers. The work presented shows high accuracy and promising results in aiding safe overtaking, with significant improvements to overtaking maneuvers on two-lane rural roads.

Proactive longitudinal control to preclude disruptive lane changes of human-driven vehicles in mixed-flow traffic

Connected and autonomous vehicles (CAVs) can be leveraged to enable cooperative platooning control to alleviate traffic oscillations. However, prior to a pure CAV environment, CAVs and human-driven vehicles (HDVs) will coexist on roads, creating a mixed-flow traffic environment. Mixed-flow traffic introduces key challenges for CAV operations due to potential lane changes by HDVs in adjacent lanes, which can cause stop-and-go waves and traffic oscillations. An understanding of the interactions between CAVs and HDVs in the lane-change process can be leveraged to use CAVs to proactively preclude disruptive lane changes by HDVs. This study proposes a deep reinforcement learning-based proactive longitudinal control strategy (PLCS) for CAVs to counteract disruptive HDV lane-change behaviors that can induce disturbances, and to preserve the smoothness of traffic flow in the CAV platooning control process. In it, a Transformer-based lane-change traffic condition predictor is constructed to predict whether an HDV will likely perform a disruptive lane change under the ambient traffic conditions. If no disruptive lane change is predicted, an extended intelligent driver model is activated for the CAV to perform smooth car-following behavior under cooperative CAV platooning control. If a disruptive lane change is predicted, a rainbow deep Q-network (RDQN)-based lane-change preclusion model is proposed through which the CAV can alter the lane-change traffic condition to preclude the HDV’s lane change. Results from numerical experiments suggest that a CAV controlled by the PLCS is effective in reducing disruptive lane-change maneuvers by an HDV in its vicinity, and can improve string stability performance in mixed-flow traffic. Further, the effectiveness of the PLCS is illustrated under different lane-change scenarios, CAV control setups, and HDV driver types.

Risk-aware controller for autonomous vehicles using model-based collision prediction and reinforcement learning

Autonomous Vehicles (AVs) have the potential to save millions of lives and increase the efficiency of transportation services. However, the successful deployment of AVs requires tackling multiple challenges related to modeling and certifying safety. State-of-the-art decision-making methods usually rely on end-to-end learning or imitation learning approaches, which still pose significant safety risks. Hence the necessity of risk-aware AVs that can better predict and handle dangerous situations. Furthermore, current approaches tend to lack explainability due to their reliance on end-to-end Deep Learning, where significant causal relationships are not guaranteed to be learned from data.This paper introduces a novel risk-aware framework for training AV agents using a bespoke collision prediction model and Reinforcement Learning (RL). The collision prediction model is based on Gaussian Processes and vehicle dynamics, and is used to generate the RL state vector. Using an explicit risk model increases the post-hoc explainability of the AV agent, which is vital for reaching and certifying the high safety levels required for AVs and other safety-sensitive applications. Experimental results obtained with a simulator and state-of-the-art RL algorithms show that the risk-aware RL framework decreases average collision rates by 15%, makes AVs more robust to sudden harsh braking situations, and achieves better performance in both safety and speed when compared to a standard rule-based method (the Intelligent Driver Model). Moreover, the proposed collision prediction model outperforms other models in the literature.

Longitudinal car-following control strategy integrating predictive collision risk

Many controllers have been designed for the longitudinal car-following mode of connected and automated vehicles. The existing body of research shows that these controllers more or less have some deficiencies in controlling connected and automated vehicles, which directly hinder their development. To cope with these problems, we design a novel longitudinal car-following control strategy integrating predictive collision risk. Among the controllers for connected and automated vehicles, the intelligent driver model is one of the most widely adopted controller. There is evidence that the intelligent driver model controller will produce larger gap distance and slower response time. In the meantime, the string stability of CAVs platoon controlled by the intelligent driver model controller is poor. Therefore, to verify the effectiveness of this control strategy, we integrate proposed control strategy with intelligent driver model, and the predictive-intelligent drive model controller is formulated. Theoretical analysis is exerted by transfer function method, and some numerical simulation experiments are designed and carried out. Analytical results show that the proposed control strategy can improve the defects of intelligent driver model controller in poor string stability, larger gap distance and slower response. In the meantime, it can also moderate the stopping process of connected and automated vehicles and enhance the throughput of intersection. More importantly, it does not affect the structure of connected and automated vehicles, which suggests that it is possible to integrate it with other controllers to improve their performance in controlling connected and automated vehicles. All findings in this study provide a novel way to establish car-following control strategy improving traffic stability and efficiency and a set of theoretical frameworks for verifying the control effect of future proposed control strategy.

Cooperative predictive control for arbitrarily mixed vehicle platoons with guaranteed global optimality

AbstractThis paper studies the cooperative control problem of a mixed vehicle platoon, which is composed of connected autonomous vehicles (CAVs) and human‐driven vehicles (HDVs) in an arbitrary order. An alternating direction method of multipliers (ADMM) based distributed model predictive control (DMPC) algorithm is proposed for CAVs to lead the mixed vehicle platoon travelling in a string with anticipated inter‐vehicle spacing and a desired velocity. First, the mixed vehicle platoon is divided into multiple interrelated sub‐platoons with any two adjacent sub‐platoons having a common CAV, and then a generic model is constructed for each sub‐platoon based on the intelligent driver model for HDV and the kinematic model for CAV, respectively. Second, a local MPC controller is designed for each sub‐platoon to optimize the control inputs of CAVs with the objective of minimizing the position and velocity errors of all vehicles in the sub‐platoon, and then the ADMM is utilized to obtain the global optimal solution among local MPC controllers of all sub‐platoons. Finally, numerical simulations and experiments are carried out to verify the effectiveness of the proposed DMPC algorithm, the results of which reveal that it can reduce the computation cost significantly and ensure the control performance comparable to the centralized MPC algorithm.

Consistency Analysis of Drivers’ Car-Following Behaviors

Car-following models are calibrated to account for various driver behaviors such as speed and space headway. Because drivers do not all drive the same way, they are typically classified based on their level, or profile, of aggressiveness. This approach to model calibration assumes that a single set of car-following parameters applies to an individual driver consistently, that is, during the entire driving period a study considers. The purpose of this research is to challenge that assumption by analyzing the heterogeneity and inconsistency of driver behaviors in different traffic conditions. A total of 3,262 urban expressway car-following periods from 51 drivers were extracted from the Shanghai Naturalistic Driving Study database. The intelligent driver model (IDM) was selected as this study’s car-following model for its favorable performance in highly dynamic situations; its parameters, that is, desired speed, desired time headway, maximum acceleration, comfortable deceleration, acceleration exponent, and minimum spacing, were calibrated. To reflect the effect of traffic conditions, the car-following periods were categorized into three regimes from low-speed to high-speed. Two-way analysis of variance was used to evaluate heterogeneity and inconsistency in the car-following parameters. The findings of this study show that (1) the main significant variable that distinguished driver profile was desired time headway, while the comfortable acceleration was commonly affected by both driver profile and speed regime; (2) variability of individual driver parameter estimates greatly depended on the traffic regime (low-, medium-, and high-speed), showing that driver behavior was highly influenced by traffic conditions; and (3) speed information was demonstrated to be a useful tool by which to divide car-following periods for modeling, as the car-following parameters exhibited distinct differences among the regimes in the varied traffic environment.