Driver Agent Research Articles

Autonomous agent adaptive driving control based on fuzzy logic theory and normative behavior

Studying driver behaviors has become a major concern for the transportation community, businesses, and the public. Thus, based on the simulation, we proposed an adaptive driving model in the car-following driving behavior and based on the normative behavior of the driver during decision-making and anticipation, whose intention is to ensure the objectives of imitation of ordinary human behavior and road safety. The presented model is based on a software agent paradigm to model a human driver and the Fuzzy Logic Theory to reflect the driver agent’s reasoning. To validate our model, we used the dataset from the program of the US Federal Highway Administration. In this context, we notice an excellent homogeneity in the deviation of the adopted trajectory of the autonomous driver agent from the adopted trajectories by the human drivers. Moreover, the advantage of our model is that it works with different velocities.

A Reinforcement Learning Approach for Enacting Cautious Behaviours in Autonomous Driving System: Safe Speed Choice in the Interaction With Distracted Pedestrians

Driving requires the ability to handle unpredictable situations. Since it is not always possible to predict an impending danger, a good driver should preventively assess whether a situation has risks and adopt a safe behavior. Considering, in particular, the possibility of a pedestrian suddenly crossing the road, a prudent driver should limit the traveling speed. We present a work exploiting reinforcement learning to learn a function that specifies the safe speed limit for a given artificial driver agent. The safe speed function acts as a behavioral directive for the agent, thus extending its cognitive abilities. We consider scenarios where the vehicle interacts with a distracted pedestrian that might cross the road in hard-to-predict ways and propose a neural network mapping the pedestrian&#x2019;s context onto the appropriate traveling speed so that the autonomous vehicle can successfully perform emergency braking maneuvers. We discuss the advantages of developing a specialized neural network extension on top of an already functioning autonomous driving system, removing the burden of learning to drive from scratch while focusing on learning safe behavior at a high-level. We demonstrate how the safe speed function can be learned in simulation and then transferred into a real vehicle. We include a statistical analysis of the network&#x2019;s improvements compared to the original autonomous driving system. The code implementing the presented network is available at <uri>https://github.com/tonegas/safe-speed-neural-network</uri> with MIT license and at <uri>https://zenodo.org/communities/dreams4cars</uri>.

A Bayesian Driver Agent Model for Autonomous Vehicles System Based on Knowledge-Aware and Real-Time Data

A key research area in autonomous driving is how to model the driver’s decision-making behavior, due to the fact it is significant for a self-driving vehicles considering their traffic safety and efficiency. However, the uncertain characteristics of vehicle and pedestrian trajectories affect urban roads, which poses severe challenges to the cognitive understanding and decision-making of autonomous vehicle systems in terms of accuracy and robustness. To overcome the abovementioned problems, this paper proposes a Bayesian driver agent (BDA) model which is a vision-based autonomous vehicle system with learning and inference methods inspired by human driver’s cognitive psychology. Different from the end-to-end learning method and traditional rule-based methods, our approach breaks the driving system up into a scene recognition module and a decision inference module. The perception module, which is based on a multi-task learning neural network (CNN), takes a driver’s-view image as its input and predicts the traffic scene’s feature values. The decision module based on dynamic Bayesian network (DBN) then makes an inferred decision using the traffic scene’s feature values. To explore the validity of the Bayesian driver agent model, we performed experiments on a driving simulation platform. The BDA model can extract the scene feature values effectively and predict the probability distribution of the human driver’s decision-making process accurately based on inference. We take the lane changing scenario as an example to verify the model, the intraclass correlation coefficient (ICC) correlation between the BDA model and human driver’s decision process reached 0.984. This work suggests a research in scene perception and autonomous decision-making that may apply to autonomous vehicle system.

Comparative Study of End-to-end Deep Learning Methods for Self-driving Car

Self-driving car is one of the most amazing applications and most active research of artificial intelligence. It uses end-to-end deep learning models to take orientation and speed decisions, using mainly Convolutional Neural Networks for computer vision, plugged to a fully connected network to output control commands. In this paper, we introduce the Self-driving car domain and the CARLA simulation environment with a focus on the lane-keeping task, then we present the two main end-to-end models, used to solve this problematic, beginning by Deep imitation learning (IL) and specifically the Conditional Imitation Learning (COIL) algorithm, that learns through expert labeled demonstrations, trying to mimic their behaviors, and thereafter, describing Deep Reinforcement Learning (DRL), and precisely DQN and DDPG (respectively Deep Q learning and deep deterministic policy gradient), that uses the concepts of learning by trial and error, while adopting the Markovian decision processes (MDP), to get the best policy for the driver agent. In the last chapter, we compare the two algorithms IL and DRL based on a new approach, with metrics used in deep learning (Loss during training phase) and Self-driving car (the episode's duration before a crash and Average distance from the road center during the testing phase). The results of the training and testing on CARLA simulator reveals that the IL algorithm performs better than DRL algorithm when the agents are already trained on a given circuit, but DRL agents show better adaptability when they are on new roads.

Multi-Agent Cooperation for an Active Perception Based on Driving Behavior: Application in a Car-Following Behavior

ABSTRACT Perception is presented as a predominant concern in the functioning of a driving system, where it is necessary to understand how the information, events, and actions of each influence the state of the environment and the objectives of the driver, immediately and in the near future. In this context, we present in this paper a driving model composed of five layers which ensure the autonomy and road safety of a driver agent, in particular, we are interested in this article in the concept of perception which is translated by the first three layers of our driving model, which are: visual perception, comprehension and projection, where the execution of these three layers is based on the driving behavior adopted by the driver agent, which is in our case the car-following driving behavior. Furthermore, we present in this paper two simulation scenarios, the first one is realized based on urban area conditions, and the second one is conducted by using Next Generation SIMulation (NGSIM) dataset of a highway in Los Angeles, California. In this context, the experimental results present the effectiveness of our driving model based on the imitation of human behavior and according to reducing the duration of perception.

Preliminary Study for Feasibility of Driver Agent in Actual Car Environment&amp;lt;br/&amp;gt;—Driver Agent for Encouraging Safe Driving Behavior (3)

Japan has become a more aged society and there are more drivers, 65 years of age and above. Cars represent an important mode of transportation for the elderly; however, in recent years, the number of traffic accidents caused by elderly drivers has been on the rise, and this has become a social issue. Thus, for the elderly drivers to encourage them to improve their driving, we study a driver agent system which consists of smartphone, communication robot and cloud service and provides the driving support by attention awakening and the feedback support based on driving behavior evaluation. In this paper, we presented a summary of the proposed agent and reported on a set of preliminary experiments using our agent in an actual car environment. From the analysis of subjective evaluations and fixation points during driving, the results revealed the possibility that the drivers accept the agent and supports from the agent during driving and that the agent in an actual car environment did not distract the driver.

A literature review of mobility services: definitions, modelling state-of-the-art, and key considerations for a conceptual modelling framework

ABSTRACTThis paper presents a comprehensive literature review focused on the supply side of mobility services, providing relevant insights at the conceptual, operational, and modelling levels. Definitions are first drawn from the Mobility as a Service paradigm due to its predominance in the literature. This is followed by an assessment of the operational features of a range of mobility services, including carsharing, bikesharing, ridehailing, and demand responsive transit. To conclude the review, the state-of-the-art in modelling approaches for mobility services is reported, at different levels of complexity and integration. Three of the most important findings and arguments from this paper suggest that a high degree of generality exists for operational features of mobility services; that it is essential to make a distinction between Mobility as a Service and a mobility service in isolation; along with the argument that human agency should be carefully considered in modelling efforts, both for user agent and driver agent decision-making processes. Finally, key considerations are proposed for the future development of a conceptual framework for modelling the supply side of mobility services, which would have a generic service provider model as its core component.

On Control of Multiagent Systems in the Presence of a Misbehaving Agent

We focus on multiagent systems that involve agents exchanging local information through a fixed, connected, and undirected graph. In the presence of an agent subject to a disturbance (i.e., misbehaving agent), we investigate the stability and the steady-state value of the multiagent system when a proportional-integral controller is applied to a selected agent only (i.e., the driver agent) due to limited resources. Specifically, we first show that the employment of the proportional-integral controller in the driver agent makes the resulting closed-loop system matrix Hurwitz. We next derive the steady-state value of the multiagent system when the controller is respectively applied to an undisturbed agent and the misbehaving agent. For fixed, connected, and undirected acyclic graphs, we finally utilize graph-theoretic tools in order to explicitly calculate the steady-state values of all agents. To this end, we characterize several properties such as the minimization of the largest steady-state error in the multiagent system and the maximization of the number of agents having zero steady-state error.

Bi-level Decision-making Modeling for an Autonomous Driver Agent: Application in the Car-following Driving Behavior

ABSTRACTRoad crashes are present as an epidemic in road traffic and continue to grow up, where, according to World Health Organization; they cause more than 1.24 million deaths each year and 20 to 50 million non-fatal injuries, so they should represent by 2020 the third leading global cause of illness and injury. In this context, we are interested in this paper to the car-following driving behavior problem, since it alone accounts for almost 70% of road accidents, which they are caused by the incorrect judgment of the driver to keep a safe distance. Thus, we propose in this paper a decision-making model based on bi-level modeling, whose objective is to ensure the integration between road safety and the reducing travel time. To ensure this objective, we used the fuzzy logic approach to model the anticipation concept in order to extract more unobservable data from the road environment. Furthermore, we used the fuzzy logic approach in order to model the driver behaviors, in particular, the normative behaviors. The experimental results indicate that the decision to increase in velocity based on our model is ensured in the context of respecting the road safety.

Study on Driver Agent Based on Analysis of Driving Instruction Data — Driver Agent for Encouraging Safe Driving Behavior (1) —

Study on Driver Agent Based on Analysis of Driving Instruction Data — Driver Agent for Encouraging Safe Driving Behavior (1) —

Effect of Difference in Form of Driving Support Agent to Driver’s Acceptability&amp;lt;br/&amp;gt;—Driver Agent for Encouraging Safe Driving Behavior (2)

In recent years, the number of traffic accidents caused by elderly drivers has increased in Japan. However, a car is an important mode of transportation for the elderly. Therefore, to ensure safe driving, a system that can assist elderly drivers is required. In this study, we propose a driver-agent system that provides support to elderly drivers during and after driving and encourages them to improve their driving. This paper describes the prototype system based on the analysis of the teaching records of a human instructor, and the subjective evaluation of driving support to elderly and non-elderly driver from three different agent forms, a voice, visual, and robot. The result revealed that the robot form is more noticeable, familiar, and acceptable to the elderly and non-elderly than other forms.

Incorporation of the Driver’s Personality Profile in an Agent Model

Urban traffic flow is a complex system. Behavior of an individual driver can have butterfly effect which can become root cause of an emergent phenomenon such as congestion or accident. Interaction of drivers with each other and the surrounding environment forms the dynamics of traffic flow. Hence global effects of traffic flow depend upon the behavior of each individual driver. Due to several applications of driver models in serious games, urban traffic planning and simulations, study of a realistic driver model is important. Hhence cognitive models of a driver agent are required. In order to address this challenge concepts from cognitive science and psychology are employed to design a computational model of driver cognition which is capable of incorporating law abidance and social norms using big five personality profile.

An agent-based approach for road pricing: system-level performance and implications for drivers

Background: Road pricing is a useful mechanism to align private utility of drivers with a system-level measure of performance. Traffic simulation can be used to predict the impact of road pricing policies. The simulation is not a trivial task because traffic is a social system composed of different interacting entities. To tackle this complexity, agent-based approaches can be employed to model the behavior of the several actors in transportation systems. Methods: We model traffic as a multiagent system in which link manager agents employ a reinforcement learning scheme to determine road pricing policies in a road network. Drivers who traverse the road network are cost-minimizer agents with local information and different preferences regarding travel time and credits expenditure. Results: The vehicular flow achieved by our reinforcement learning approach for road pricing is close to a method where drivers have global information of the road network status to choose their routes. Our approach reaches its peak performance faster than a fixed pricing approach. Moreover, drivers’ welfare is greater when the variability of their preferences regarding minimization of travel time or credits expenditure is higher. Conclusions: Our experiments showed that the adoption of reinforcement learning for determining road pricing policies is a promising approach, even with limitations in the driver agent and link manager models.

Society of Mind cognitive agent architecture applied to drivers adapting in a traffic context

This paper challenges the idea of creating an architectural approach for designing a complex hybrid cognitive agent, i.e., an agent that possesses all reactive, deliberative and affective capabilities, outside the hierarchical paradigm. Thus, it presents an agent architecture based on Marvin Minsky’s ‘Society of Mind’ (SoM) metaphor on human mind and cognition, and shows how SoM agents, through the features implicitly generated by their design, allow a tremendous power of representation of human heterogeneity and variety of behaviours. In order to show that, the SoM general architecture is instantiated and evaluated in a cognitively demanding environment: as a driver agent in a traffic behaviour context.

Performance of heterogeneous robot teams with personality adjusted learning

Abstract This paper presents a reinforcement learning algorithm, which is inspired by human team dynamics, for autonomous robotic multi agent applications. Individual agents on the team have heterogeneous capabilities and responsibilities. The learning algorithm assigns strictly local credit assignments to individual agents promoting scalability of the team size. The Personality Adjusted Learner (PAL) algorithm is applied to heterogeneous teams of robots with reward adjustments modified from earlier work on homogeneous teams and an information-based action personality type assignment algorithm has been incorporated. The PAL algorithm was tested in a robot combat scenario against both static and learning opponent teams. The PAL team studied included distinct commander, driver, and gunner agents for each robot. The personality preferences for each agent were varied systematically to uncover team performance sensitivities to agent personality preference assignments. The results show a significant sensitivity for the commander agent. This agent selected the robot strategy, and it was noted that the better performing commander personalities were linked to team oriented actions, rather than more selfish strategies. The driver and gunner agent performance remained insensitive to personality assignment. The driver and gunner actions did not apply at the strategic level, indicating that personality preferences may be important for agents responsible for learning to cooperate intentionally with teammates.

Decision support for simulating the car park activity in an urban area

AbstractComplexity of car park activity is reproduced from a concurrent execution of behaviour of various drivers. This paper presents a step in the development of a multimodal traffic simulator based on multi‐agent paradigm and designed as a decision aid tool as well as a video game. The user‐player has the opportunity to test different scenarios. We propose an approach for designing the decision‐making rules and the learning mechanism for a car driver agent. For that, a panel of methods such as stated preference modelling, Design Of Experiments and data fusion is used. Initial behavioural models, based on similar preferences, are developed for specified categories. Each agent will adapt its behaviour after executing its learning process. Our approach can be used in order to optimize needs of road network users and those of people in charge of traffic regulation. A demonstrator has been developed to test parking policies in an urban area as well as changes of car park characteristics.

Evaluating Urban Parking Policies with Agent-Based Model of Driver Parking Behavior

This paper presents an explicit agent-based model of parking search in a city. In the model, “drivers” drive toward their destination, search for parking, park, remain at the parking place, and leave. The city's infrastructure is represented by a high-resolution geographic information system (GIS) of the street network and parking lots; information is included on traffic directions and permitted turns, on-street parking permissions, and layers of off-street parking places and lots. Destinations are presented by layers of dwellings and public places. Driver agents belong to one of four categories: residents and guests with dwellings as destinations and employees and customers with public places as destinations. Each agent has its own destination, willingness to pay, time of arrival, and duration of stay. In the model, driver agents are “landed” at a distance of approximately 250 m from their destination, that is, close to the area in which drivers start searching for parking. First, a driver estimates the parking situation in the area and then starts to search for a parking place. During the search, a driver agent accounts for the availability of parking places, differences in pricing, and parking enforcement efforts. The model outputs include distributions of (a) search time, (b) distance between parking place and destination, (c) fees paid by the drivers, and (d) parking revenues for the proprietor of paid parking places (whether local authority or private operator). The model is implemented as an ArcGIS application and applied to analyze parking dynamics in an inner city neighborhood in Tel Aviv, Israel, during the course of a regular weekday.

Traffic Flow Simulator by Multiagent

Several methods have been proposed for investigating complex traffic flows. However, the dynamics of vehicles and their drivers' characteristics, which are important factors for any traffic flow analysis, have not been considered sufficiently enough in the past. This paper describes a new concept for simulating complex traffic flows by multiple agents. The agents are considered as a kind of artificial life (A-Life), and their behaviors are independently based on their own knowledge. Drivers, vehicles, roads, and the traffic environment, which are essential elements in our simulation, are defined as the agents. The combination of the driver agent and the vehicle agent work based on the drivers' fields of vision and their unique individual characteristics. The road agent communicates with the vehicle agent in regard to its velocity and position. The effectiveness of our new method for modeling a traffic flow is shown by a simulation study.

Spontaneous Ignition of an Inflammable Gaseous Mixture on Release from Compression

THE ‘shock-tube’ method of producing one-dimensional shock waves has found many applications in recent years. It was first used by Vieille1 to study the establishment and propagation of discontinuous waves in columns of air. Payman and Shepherd2 applied the method—in their ‘bursting-diaphragm apparatus’—to examine, by means of schlieren photography, the nature and spread of the discontinuous disturbance in the expansion chamber. Their results3 linking shock-wave speed and air flow with the initial pressure in the compression chamber fitted accepted theory. The theory of shock tubes has since been more deeply considered, for example, by Hertzberg4 in the United States, and by Lukasiewicz5 in Canada. A shock tube has also been used by Perry and Kantrowitz6 for demonstrating the luminosity caused by an intense shock wave being propagated in air. Recent developments have been directed towards the use of the shock tube for the generation of shock waves of high Mach number, and for this purpose gases of high acoustic speed have been used in the compression end as the driver agent, the high speed and pressure being obtained by exploding inflammable mixtures of low density behind the diaphragm. Hertzberg and Smith7 describe some anomalous results in such experiments, and it is to offer perhaps an explanation of these that this communication is made.