Driving Simulator Research Articles

The Analysis of Driver’s Recognition Time of Different Traffic Sign Combinations on Urban Roads via Driving Simulation

Given the impact of traffic sign combinations (TSC) on the driver’s visual recognition, this paper analyzed the influence on the driver’s visual recognition process. It used the cognitive psychology theory to establish the information transmission model during the traffic sign combinations. It abstracted the information transmission model to construct the driver’s information processing model. Simultaneously, according to the analysis of the traffic sign combinations of the urban roads, this paper carried out the driver’s visual recognition simulation test when the traffic signs were combined, measured the reaction time of the driver’s visual recognition of multiple combinations of traffic signs, and analyzed the driver’s recognition time (DRT) range in the traffic sign combinations. It used correlation analysis, robust estimation, polynomial regression, and other methods to obtain a significant relationship between the driver’s recognition times in different traffic sign combinations (DTSC). Then it built polynomial regression analysis models, fitted the data, and visualized the fitting results. The results show that through the analysis of the experimental data, based on ensuring certain accuracy, the driver’s recognition time of the traffic sign combinations of the urban road increased appropriately. There is a significant relationship between different traffic sign combinations and the driver’s recognition time. As the number of traffic signs increases, the driver’s recognition time increases significantly. Besides, under certain conditions, gender, age, and the driving experience will impact the driver’s recognition time during the traffic sign combinations. The research results can provide the relevant theoretical basis for the setting of urban traffic signs, provide a powerful reference for the revision of various traffic sign setting standards and norms, and provide ideas for future research on the sign system.

Quasi-LPV Interconnected Observer Design for Full Vehicle Dynamics Estimation With Hardware Experiments

Safety systems claim an in-depth study of vehicle motion and tire-ground interaction for the design of a partially automated driving vehicle. This article addresses the quasi-linear parameter-varying Luenberger interconnected fuzzy observer to estimate both longitudinal and lateral vehicle dynamics simultaneously. In a different manner from the commonplace state of the art of vehicle state observers that consider single driving motion, the proposed approach considers the coupled dynamics with the tire-ground interaction to estimate the most important states while reducing the complexity related to the observability and conservatism. This consideration leads to a nonlinear parameter-dependent interconnected model with unmeasured premise variables. Then, the Takagi-Sugeno fuzzy form is considered to deal with the nonlinearities of the vehicle longitudinal and lateral speeds and the slip velocities as well as the steering angle. The concept of “input to state stability” is exploited using fuzzy nonquadratic Lyapunov stability arguments to guarantee the boundness of the estimation errors. A refinement has been proposed through the so-called Polya's theorem aiming to further reduce the conservativeness. Finally, performances and effectiveness of the suggested approach are evaluated through hardware experiments performed with the well-known Sherpa dynamic car simulator under real-world driving situations.

Assessing the Effect of a Visual Navigational System on Route-Learning From an Ecological Perspective.

Route-learning, considered from an ecological approach to perception, is posited to involve the detection of information over time that specifies a path from one location to another. The study examines whether the use of a visual navigational system (e.g., GPS) may impede route-learning by drawing attention away from transitions along a path that serve as information for way-finding. Virtual reality (VR) technology used in conjunction with an extensive, detailed environmental simulation was employed to explore this possibility. One group of participants drove a simulated car in VR along a designated path while relying on visual GPS guidance. It was expected that use of the GPS display would draw attention away from temporally continuous path information. A second group initially drove the same route without GPS guidance. Both groups drove the path a second time without navigational assistance. Overall, the percentage of correct actions taken at intersections (transitions) during the second trial were significantly lower for the first group who initially drove the route with visual GPS guidance as compared to those who initially traveled the route without it. The results are consistent with the kind of trade-off that is commonplace when tools are used to mediate and assist skilled action.

Identification of drivers’ driving habits and shift schedule correction for vehicles with automatic transmission

Resource conservation has become a hot topic, and driving habits also have a significant impact on a car’s fuel consumption. In view of the different needs of drivers with different driving habits for automatic transmission and shift characteristics, a vehicle automatic transmission, and shift correction control strategy based on driving habit recognition is proposed. Based on the analysis of drivers’ driving behavior, the phase space reconstruction method is first used to reconstruct the time series of driving control signals, and the driving habit identification and gear shift correction control are conducted based on the correlation dimension and Kolmogorov entropy evaluation index. Simulation and real car test show that the identification method based on phase space reconstruction method and driving habits evaluation index can accurately identify drivers’ driving habits. The gear shift correction control strategy based on driving habit recognition fully meets the different requirements of different drivers for the gearshift performance of vehicles and improves the intelligence degree of automatic transmission of vehicles.

Haptic Devices Based on Real-Time Dynamic Models of Multibody Systems

Multibody modeling of mechanical systems can be applied to various applications. Human-in-the-loop interfaces represent a growing research field, for which increasingly more devices include a dynamic multibody model to emulate the system physics in real-time. In this scope, reliable and highly dynamic sensors, to both validate those models and to measure in real-time the physical system behavior, have become crucial. In this paper, a multibody modeling approach in relative coordinates is proposed, based on symbolic equations of the physical system. The model is running in a ROS environment, which interacts with sensors and actuators. Two real-time applications with haptic feedback are presented: a piano key and a car simulator. In the present work, several sensors are used to characterize and validate the multibody model, but also to measure the system kinematics and dynamics within the human-in-the-loop process, and to ultimately validate the haptic device behavior. Experimental results for both developed devices confirm the interest of an embedded multibody model to enhance the haptic feedback performances. Besides, model parameters variations during the experiments illustrate the infinite possibilities that such model-based configurable haptic devices can offer.

Super-Human Performance in Gran Turismo Sport Using Deep Reinforcement Learning

Autonomous car racing is a major challenge in robotics. It raises fundamental problems for classical approaches such as planning minimum-time trajectories under uncertain dynamics and controlling the car at the limits of its handling. Besides, the requirement of minimizing the lap time, which is a sparse objective, and the difficulty of collecting training data from human experts have also hindered researchers from directly applying learning-based approaches to solve the problem. In the present work, we propose a learning-based system for autonomous car racing by leveraging a high-fidelity physical car simulation, a course-progress proxy reward, and deep reinforcement learning. We deploy our system in Gran Turismo Sport, a world-leading car simulator known for its realistic physics simulation of different race cars and tracks, which is even used to recruit human race car drivers. Our trained policy achieves autonomous racing performance that goes beyond what had been achieved so far by the built-in AI, and, at the same time, outperforms the fastest driver in a dataset of over 50,000 human players.

The Effects of Image Augmentation on Efficiency of a Convolutional Neural Network of a Self-Driving Car

The objective of this paper was to study the effectiveness of image augmentation techniques in training a Convolutional Neural Network (CNN) of a self-driving car and identify the most suitable form of image augmentation technique, using the Udacity Car Simulator. Firstly, a dataset of augmented and non-augmented images from a training track, consisting of left-, right-, and front-facing views from the car cameras was created. Various image augmentation techniques were used: zoom, brightness, pan, flip, random (augments the image by arbitrarily choosing a technique from the previous four), and no augmentation. Secondly, training datasets consisting of the aforementioned images and a log of car turning angles, throttle, and brake were built. The final training datasets were then used with NVIDIA training method to train different CNN. The different trained networks generated steering commands from the front-facing camera of the simulation and test track had no effect on the generalization of the CNN. Lastly, different trained networks were used on the test track of Udacity Car Simulator to calculate the following variables: distance travelled, and number of crashes made by the car. After these values were acquired, an efficiency analysis was performed. The results suggested augmentation of training data is a crucial factor when it comes to the process of generalizing a model to perform tasks. Random augmentations performed the best, but a combination of flip and brightness augmentations performed equally efficiently.

End-to-End Autonomous Driving Through Dueling Double Deep Q-Network

Recent years have seen the rapid development of autonomous driving systems, which are typically designed in a hierarchical architecture or an end-to-end architecture. The hierarchical architecture is always complicated and hard to design, while the end-to-end architecture is more promising due to its simple structure. This paper puts forward an end-to-end autonomous driving method through a deep reinforcement learning algorithm Dueling Double Deep Q-Network, making it possible for the vehicle to learn end-to-end driving by itself. This paper firstly proposes an architecture for the end-to-end lane-keeping task. Unlike the traditional image-only state space, the presented state space is composed of both camera images and vehicle motion information. Then corresponding dueling neural network structure is introduced, which reduces the variance and improves sampling efficiency. Thirdly, the proposed method is applied to The Open Racing Car Simulator (TORCS) to demonstrate its great performance, where it surpasses human drivers. Finally, the saliency map of the neural network is visualized, which indicates the trained network drives by observing the lane lines. A video for the presented work is available online, https://youtu.be/76ciJmIHMD8 or https://v.youku.com/v_show/id_XNDM4ODc0MTM4NA==.html.

AmbientBreath

To promote calm breathing inside a car, we designed a just-in-time breathing intervention stimulated by multi-sensory feedback and evaluated its efficacy in a driving simulator. Efficacy was measured via reduction in breathing rate as well as by user acceptance and driving safety measures. Drivers were first exposed to demonstrations of three kinds of ambient feedback designed to stimulate a goal breathing rate: (1) auditory (rhythmic background noise), (2) synchronized modulation of wind (dashboard fans modulating air pointed toward the driver) together with auditory, or (3) synchronized visual (ambient lights) together with auditory. After choosing one preference from these three, each driver engaged in a challenging driving task in a car simulator, where the ambient stimulation was triggered when their breathing exceeded a goal rate adapted to their personal baseline. Two user studies were conducted in a car simulator involving respectively 23 and 31 participants. The studies include both manual and autonomous driving scenarios to evaluate drivers' engagement in the intervention under different cognitive loads. The most frequently selected stimulation was the combined auditory and wind modalities. Measures of changes in breathing rate show that the participants were able to successfully engage in the breathing intervention; however, several factors from the driving context appear to have an impact on when the intervention is or is not effective.

Deep Deterministic Policy Gradient Algorithm Based on Convolutional Block Attention for Autonomous Driving

The research on autonomous driving based on deep reinforcement learning algorithms is a research hotspot. Traditional autonomous driving requires human involvement, and the autonomous driving algorithms based on supervised learning must be trained in advance using human experience. To deal with autonomous driving problems, this paper proposes an improved end-to-end deep deterministic policy gradient (DDPG) algorithm based on the convolutional block attention mechanism, and it is called multi-input attention prioritized deep deterministic policy gradient algorithm (MAPDDPG). Both the actor network and the critic network of the model have the same structure with symmetry. Meanwhile, the attention mechanism is introduced to help the vehicles focus on useful environmental information. The experiments are conducted in the open racing car simulator (TORCS)and the results of five experiment runs on the test tracks are averaged to obtain the final result. Compared with the state-of-the-art algorithm, the maximum reward increases from 62,207 to 116,347, and the average speed increases from 135 km/h to 193 km/h, while the number of success episodes to complete a circle increases from 96 to 147. Also, the variance of the distance from the vehicle to the center of the road is compared, and the result indicates that the variance of the DDPG is 0.6 m while that of the MAPDDPG is only 0.2 m. The above results indicate that the proposed MAPDDPG achieves excellent performance.

Optimizing hyperparameters of deep reinforcement learning for autonomous driving based on whale optimization algorithm.

Deep Reinforcement Learning (DRL) enables agents to make decisions based on a well-designed reward function that suites a particular environment without any prior knowledge related to a given environment. The adaptation of hyperparameters has a great impact on the overall learning process and the learning processing times. Hyperparameters should be accurately estimated while training DRL algorithms, which is one of the key challenges that we attempt to address. This paper employs a swarm-based optimization algorithm, namely the Whale Optimization Algorithm (WOA), for optimizing the hyperparameters of the Deep Deterministic Policy Gradient (DDPG) algorithm to achieve the optimum control strategy in an autonomous driving control problem. DDPG is capable of handling complex environments, which contain continuous spaces for actions. To evaluate the proposed algorithm, the Open Racing Car Simulator (TORCS), a realistic autonomous driving simulation environment, was chosen to its ease of design and implementation. Using TORCS, the DDPG agent with optimized hyperparameters was compared with a DDPG agent with reference hyperparameters. The experimental results showed that the DDPG’s hyperparameters optimization leads to maximizing the total rewards, along with testing episodes and maintaining a stable driving policy.

Measurement and Comparison Study of Road Roughness Conditions by using Two Compact Road Profiler for Evaluating Ride Quality

More recently, it is strongly required that road management involves a wide range of topics over the world and it must be considered from a perspective of customer-oriented levels of service and road administrators are requested to present a visual map of road network in terms of road network monitoring among others. To fulfill these requirements, firstly, this paper introduces new topics, which is related to two Compact Road Profiler (CRP) to monitor the road roughness conditions including localized distresses. In addition, we analyze road profile data considering two wheel paths by the survey vehicle, known as inner and outer wheel paths, and using a sophisticated driving simulator (DS) to evaluate ride quality conditions based on Power Spectrum Density (PSD) function to identify the sensitive range of ride quality. Finally, we calculate the Root Mean Square (R.M.S) and Crest Factor (CF) using vertical acceleration data of the passenger seat in DS. Consequently, it is shown that advanced evaluation of the ride quality under two wheel paths is performed. As regards standard for ride evaluation, ISO 2631 standard is adopted.

Modeling, Simulation and Analysis of Electric Vehicle Driven by Induction Motor

The electric vehicle (EV)was spread around the world due to the environmental considerations. Induction motor (IM) is one type of electric motors used as main drive for EV. The present work uses Matlab/Simulink for modeling and simulation of MITSUBISHI I-MIEV model city car. Three case studies were taken in the simulation of complete (EV+IM) model with different loads and speeds. The simulation tests indicate the ability of IM in driving the EV with high stability in different motion mode.

The Evaluation of Experimental Variables for Sustainable Virtual Road Safety Audits

This paper evaluates experimental variables for virtual road safety audits (VRSAs) through practical experiments to promote sustainable road safety. VRSAs perform road safety audits using driving simulators (DSs), and all objects in the road environment cannot be experimental variables because of realistic constraints. Therefore, the study evaluates the likelihood of recommendation of VRSA experimental variables by comparing DSs experiments and field reviews to secure sustainable road safety conditions. The net promoter score results evaluated “Tunnel”, “Bridge”, “Underpass”, “Footbridge”, “Traffic island”, “Sign”, “Lane”, “Road marking”, “Traffic light”, “Median barrier”, “Road furniture”, and “Traffic condition” as recommended variables. On the contrary, the “Road pavement”, “Drainage”, “Lighting”, “Vehicle”, “Pedestrian”, “Bicycle”, “Accident”, and “Hazard event” variables were not recommended. The study can be used for decision making in VRSA scenario development as an initial effort to evaluate its experimental variables.

A Behavioral Model of Drivers’ Mean Speed Influenced by Weather Conditions, Road Geometry, and Driver Characteristics Using a Driving Simulator Study

Driving above the speed limit is one of the factors that significantly affect safety. Many studies examined the factors affecting the speed of vehicles in the simulated environment. The present study aimed to analyze drivers’ characteristics, time and weather conditions, and geometric features’ effect on mean speed in simulated conditions simultaneously. In this regard, the simulator experiment data of 70 drivers were collected in a two-lane rural highway at six different times, and weather scenarios and their socioeconomic characteristics were collected by a questionnaire. Structural equation modeling (SEM) was used to capture the complex relationships among related variables. Eleven variables were grouped into four latent variables in the structural model. Latent variables including “Novice Drivers,” “Experienced Drivers,” “Sight Distance,” and “Geometric Design” were defined and found significant on their mean speed. The results showed that “Novice Drivers” have a positive correlation with the mean speed. Meanwhile, “Experienced Drivers,” who drive 12% slower than the novice group, negatively affect the mean speed with a standard regression weight of −0.08. This relation means that young and novice drivers are more inclined to choose higher speeds. Among variables, the latent variable “Sight Distance” has the most significant effect on the mean speed. This model shows that foggy weather conditions strongly affect the speed selection behavior and reduce the mean speed by 40%. Nighttime also reduces mean speed due to poor visibility conditions. Furthermore, “Geometric design” as the latent variable indicates the presence of curves on the simulated road, and it can be concluded that the existence of a curve on the road encourages drivers to slow down, even young drivers. It is noteworthy that the parts of the simulated road with a horizontal curve act as a speed reduction tool for drivers.

Comparative evaluation of dynamics and fuel efficiency indicators of the car with robotic stepped transmission under condition of different diameters of engine cylinders

Problem. Repair dimensions method regulates increment of cylinders by the means of mechanical treatment, which leads to increasing of working volume of the engine. This will cause a change of dynamics and fuel-efficiensy indicators of the car with the certain type of transmission. Goal. The goal is to comparatively evaluate dynamism and fuel-efficiency of automotive vehicle category N2 with two options of engines, which have nominal and repair cylinder sizes. The riding cycle of the cars with robotic stepped gearbox was researched, that is without breaking of force stream while switching gears. Methology. Imitative simulation of cars riding cycle, according to standard, was conducted with the help of Matlab Simulink software. Because of the fact that mathematical model of cars movement during mentioned cycle is based on elementary dependencies, for its realization the only basics avialable blocks of the sofware package were used. The flowchart is given in the paper. Results. Simulation of riding cycle of N2 – category automotive vehicle with robotic gearbox was conducted at full engine’s capacity while running-up. Obtained results have shown that engines with repairment sizes have advantage in dynamism indicators, engines with nominal cylinder sizes have advantages in fuel-efficiency. Originality. For obtaining of comparative evaluation of dynamics and fuel-efficiency indicators of the car with robotic gearbox with different engines, the imitation simulation of riding cycle, according to standarts, was used with the help of Matlab Simulink software. Practical value. Application of the repairment sizes method upon the engine’s cylinder liners has shown improvement of dynamics indicators of car with robotic transmission, which will improve its productivity under excessive engines work conditions.

Development of a Mobile Driving Simulator to Eliminate False Motion Cues


 
 
 Driving Simulators are a valuable tool for the evaluation of driver assistance systems and analysis of user behaviour. They consist of a vehicle mock-up and a display, motion and an audio system. As, driving is mainly a visual task and the driver receives most of the information through his eyes, so the configuration of the display is very important for accurate perception of surroundings. Important features of a display system are its distance from driver's eyes, field of view, continuity and the picture quality of the displayed image. In order to simulate motion, most of the existing driving simulators consist of a dome, mounted on a Stewart platform, which is either stationary or moves on a rail or a horizontal table. Due to the limiting working space of the motion system of such driving simulators, they cannot accurately simulate longitudinal accelerations so they use scaled vehicle dynamics model or blend the longitudinal movement of simulator with the tilt movement, which the driver perceives as unrealistic motion cues. To eliminate the issues of false motion perception in the driver, a mobile driving simulator is developed, which is to be driven on a planar area and a display system is designed around it. The display system covers the horizontal and vertical field of view of the driver and the distance of the display system from the driver's eyes is chosen in such a way that it takes into account the accommodation effects, which helps in the perception of depth. This results in a display system in the form of 220º cylindrical dome with a diameter of approximately 4.8 meter.
 
 

Stop or go decisions at the onset of yellow light in a connected environment: A hybrid approach of decision tree and panel mixed logit model

Driver decisions at the onset of yellow traffic lights are often critical, as inaccurate decisions may result in traffic conflicts and collisions. A future connected environment where vehicles can communicate with traffic lights is expected to minimize the uncertainty associated with a driver’s decision-making at signalized intersections by providing advance information related to traffic light changes. The effectiveness of such a connected environment, however, remains unexplored due to the paucity of relevant data. This study examines driver decisions at the onset of yellow traffic lights when they are assisted with advance information about traffic light changes. Seventy-eight participants with diverse backgrounds performed driving experiments on an urban route with a signalized intersection simulated in the CARRS-Q Advanced Driving Simulator. The experiment consisted of two randomized driving conditions: baseline (without advance information aids) and connected environment (with advance information aids). Contrary to the existing literature, this study employs a hybrid approach, leveraging the combined benefits of data mining to identify a priori relationships and a panel mixed logit model (more specifically, correlated grouped random parameters logit with heterogeneity-in-means approach) to account for unobserved heterogeneity as well as the correlation among random parameters. Our analysis shows that drivers in the connected environment are less likely to proceed through intersections at the onset of yellow light compared to the baseline condition. However, at the individual driver-level, the connected environment’s impact on driver decisions is mixed. Female drivers have been found to have a higher propensity for yellow light running in the connected environment than that of male drivers. Overall, the connected environment assists drivers in making safer decisions at the onset of yellow light.

A Case for Raising the Camera: A Driving Simulator Test of Camera-Monitor Systems.

This experiment provides a first-of-its-kind driving-simulator study to investigate the feasibility of camera-monitor systems (CMS) with displaced side-mounted cameras in sedans. Among the increasing number of studies investigating the replacement of side-mounted rearview mirrors with CMS, the placement of side-mounted cameras has been largely neglected. Moreover, user preferences with respect to camera placement have not been validated in a driving simulator. Past research merely has shown that the vertical camera position can affect distance perception. In a driving simulator experiment, we investigated the effects of rearward camera placement on driver acceptance and performance. Thirty-six participants performed multiple lane changes in a last safe-gap paradigm. The camera position, ego-velocity, and velocity of the approaching vehicle varied across the experiment. The results suggest a clear preference for a high rearward perspective, whereas participants disliked the lower viewpoint. However, these stark differences were only marginally mirrored in lane change performance. Average safety margins tended to decrease and their variation tended to increase for the low camera position. Even if the impact of the camera position on driving behavior seems to be small in sedans, driver expectations show clear-cut preferences. When designing CMS, this should be taken into account, as these preferences could promote the use of CMS and thus their positive impact on safety. Designers should place side-mounted cameras as high as possible to increase acceptance of CMS. Low camera positions are not recommended, as they might decrease safety margins and are not appreciated by drivers.

Hazard Perception, Presence, and Simulation Sickness-A Comparison of Desktop and Head-Mounted Display for Driving Simulation.

Driving simulators are becoming increasingly common in driver training and assessment. Since virtual reality is generally regarded as an appropriate environment for measuring risk behavior, simulators are also used to assess hazard perception, which is considered to be one of the most important skills for safe driving. Simulators, which offer challenges that are indeed comparable to driving in real traffic, but at a very low risk of physical injury, have the potential to complement theoretical and practical driver trainings and tests. Although configurations and fidelity differ considerably between driving simulators, studies comparing the impact of their distinct features on driving performance and test validity remain rare. In this context, prior research demonstrated that a wider field of view (three monitors compared to a single monitor) led to earlier speed adjustments in response to potential hazards—especially for experienced drivers. The wider field of view was assumed to cause the drivers to be more present in the virtual world, which in turn provoked more natural scanning of the road and therefore, earlier hazard detection in experienced drivers. Research on spatial presence in other contexts support this assumption. The present experiment investigated whether this effect could be enhanced by an even more immersive presentation technique for driving simulation: a head-mounted display (HMD). Moreover, we studied the interplay between display mode, sense of presence and simulation sickness. Eighty experienced and less experienced drivers completed six simulation-based hazard perception scenarios, which were displayed either via a triple-monitor set-up or an HMD. Results indicate that the experienced drivers showed very similar driving and risk behavior as the inexperienced drivers in both experimental conditions. However, there were significant differences between the two display conditions. The use of an HMD resulted in a clearer and more abrupt speed reduction, more virtual presence, and a higher degree of simulation sickness. However, the interrelation between these three variables could not be conclusively clarified in the present study and thus represents a research aim that could be addressed in future studies.