Access the full text.
Sign up today, get DeepDyve free for 14 days.
Loading next page...
/lp/unpaywall/driver-behavior-modeling-developments-and-future-directions-AaJjArItXe
References
References for this paper are not available at this time. We will be adding them shortly, thank you for your patience.
- Publisher
- Unpaywall
- ISSN
- 1687-5702
- DOI
- 10.1155/2016/6952791
- Publisher site
- See Article on Publisher Site
Abstract
Hindawi Publishing Corporation International Journal of Vehicular Technology Volume 2016, Article ID 6952791, 12 pages http://dx.doi.org/10.1155/2016/6952791 Review Article 1 2 Najah AbuAli and Hatem Abou-zeid College of Information Technology, UAE University, Al-Ain, UAE School of Computing, Queen’s University, Kingston, ON, Canada Correspondence should be addressed to Najah AbuAli; [email protected] Received 31 August 2016; Accepted 8 November 2016 Academic Editor: Abdelaziz Bensrhair Copyright © 2016 N. AbuAli and H. Abou-zeid. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. The advances in wireless communication schemes, mobile cloud and fog computing, and context-aware services boost a growing interest in the design, development, and deployment of driver behavior models for emerging applications. Despite the progressive advancements in various aspects of driver behavior modeling (DBM), only limited work can be found that reviews the growing body of literature, which only targets a subset of DBM processes. u Th s a more general review of the diverse aspects of DBM, with an emphasis on the most recent developments, is needed. In this paper, we provide an overview of advances of in-vehicle and smartphone sensing capabilities and communication and recent applications and services of DBM and emphasize research challenges and key future directions. 1. Introduction the vehicle behavior, determining the drivers’ state such as their attention level and driving competence can play a crucial Modeling driver behavior is a complex task that has garnered role in the success of ADASs. At another level, gaining significant research attention throughout the past decades. insight on the drivers’ objectives such as destination and road This interest is fueled by the constant growth of cities as preferences opens the door to novel travel assistance systems indicated by the recent statistics that the urban population and services. has grown from 746 million in 1950 to 3.9 billion in 2014 Despite the progressive advancements in various aspects (54% of the current world population) [1]. As more people of DBM, only a limited number of surveys can be found migrate to cities, the transportation infrastructure is faced that review the growing body of literature. Among those, with significant challenges leading to more accidents, deaths, lane changing models have been reviewed in [4, 5], while congestion, and environmental pollution. Studies have indi- Doshi and Trivedi [2] survey developments in driver intent cated that there are over 30,000 deaths and 1.2 million injuries prediction with emphasis on forecasting the trajectory of annually on roadways in the United States, 80% of which the vehicle in real time. Works covering driver skill and are due to driver inattention or as a result of unintended different approaches to driver models have recently been maneuvers [2, 3]. Human error is therefore the primary cause reviewed in [6]. A review of the cognitive components of of such tragedies. driver behavior can also be found in [7] where the author Driver behavior modeling (DBM) has primarily emerged addresses the situational factors and motives that influence to predict driving maneuvers, driver intent, vehicle and driver driving. The surveys mentioned above only target a subset state, and environmental factors, to improve transportation of DBMprocesses andthusamore generalreviewofthe safety and the driving experience as a whole. These models are diverseaspects of DBMwithanemphasisonthe most recent then typically incorporated into Advanced Driver Assistance developments is needed. In particular, advances in vehicle System (ADAS) in the vehicles. For instance, by coupling sensing capabilities (as well as smartphones), vehicle-to- sensing information with accurate lane changing prediction vehicle (V2V) communication, and cloud-based services are models, an ADAS can prevent accidents by warning the driver facilitating an unprecedented era of data collection that is ahead of time of potential danger. In addition to modeling enabling researchers to develop more sophisticated DBM. 2 International Journal of Vehicular Technology Sensing Applications Future systems Vehicle CAN: Lane changing rpm, turn Personal Driver assistance signal, Assistance Intersection acceleration Clouds decision-making Driver preference Sensors: Telematics GSP location, profiling services radar, Route planning gyroscopes Semiautonomous Autonomous Cameras: car following Driver fatigue Driver and cooperative recognition environment driving conditions Figure 1: Driver behavior modeling (DBM): sensing, applications, and future systems. The contemporary emergence of “big data” storage and modeling framework for the complete driving task. However, processing solutions is another technological development traditionally a typical modeling framework includes inputs that is anticipated also to drive new avenues of research and from various sensors and vehicle controllers, preprocessing exploration in DBM. As such, the objective of this survey is to algorithms to filter the data if necessary, the core predictive provideareview of therecentapplicationsand research areas models for particular tasks (these can follow the various in DBM and emphasize key future directions. We believe such levels discussed below), and feedback. An overview of various a state-of-the-art work is needed to assist those interested models that capture the dynamics between the driver, the in embarking in this evolving eld fi . To accomplish this we vehicle, and the environment is presented in [6, 8]. More organize thesurveyintothe followingsections: generically, DBM can be considered to involve (1)asensing phase, (2)areasoningphase,and (3)anapplication layer, (i) Section 2 rfi st provides an introduction to the com- as illustrated in Figure 1. eTh sensing phase involves various ponents and stages involved in driver behavior mod- forms of data collection from the vehicle, driver, and the eling, the various forms of input, and the primary environment. Thisisthentypically fedintoareasoning modeling approaches. engine with some application in mind. Although current (ii) Section 3 discusses the typical applications and uses research in individual applications has not yet matured, the of DBM with emphasis on ADASs and the emerging ultimate goal is to develop assistance systems that integrate autonomous vehicles. multiple personalized services for the driver as shown in Figure 1. This requires a high level of data abstraction and (iii) In Section 4 we review fundamental modeling objec- processing from multiple resources. tives in detail. The objectives are the specific research components that enable the development of the appli- cations mentioned earlier. This includes topics such 2.2. Inputs forDriverBehaviorModeling. Inputs to the as predicting behavior at intersections, lane changes, DBM include vehicle data from the Controller Area Network and route choice. (CAN), sensors, and more recently input from smartphones. (iv) Simulation-based and data-driven evaluation tech- niques are highlighted in Section 5. References to 2.2.1. CAN. The CAN provides accurate information of datasets for specific DBM objectives and applications several in-vehicle parameters such as the pedal positions, are provided. steering wheel angle, RPM acceleration, and turn signal state [2]. Data collected using the CAN is typically more accurate (v) Finally, Section 6 discusses several open research than that from mobile sensors as it is directly connected to issues and directions such as collaborative DBM and the vehicle. Several adapters can be used for data acquisition Driver Assistance Clouds (DACs). from the CAN such as the OBD-II (On-Board Diagnostics) Bluetoothadapter with theTorquePro Application[9]. 2. Overview and Preliminaries 2.1. Modeling Frameworks. As mentioned in Section 1, mod- 2.2.2. Sensors. Several sensor systems can be used in DBM eling driver behavior includes the driver intent, state, and such as radars, lane position sensors, Global Positioning vehicle dynamics. It is therefore difficult to develop a single System (GPS), accelerometers, and gyroscopes. eTh use of International Journal of Vehicular Technology 3 sensors embedded in smartphones is currently being inves- this modeling domain since they impact strategic maneu- tigated as an alternative/complementary input to the CAN, vers. Understanding strategic maneuvers provides additional and the outcomes of several projects have been reported context and preliminary input to tactical and operational recently [10–14]. This is particularly useful for older vehicles maneuvers by modeling the underlying driver preferences and in developing countries where smartphones are popular and long-term goals of the trip. In this regard, a lane change and may facilitate simpler integration into crowd sensing and can be modeled at the tactical level based on the strategic cloud-based services. However, sensor calibration is required input of the drivers route and behavioral information. and the accuracy may vary from device to device which is a Hatakka et al. [20, 21] have debated that the hierarchical topic of current investigation. control model would need to capture and include the driver’s general goals for life and skills for living and hence extended the three hierarchical levels into four by adding the behavioral 2.2.3. Cameras. While cameras can be considered to be level on top of the three hierarchical control levels, introduc- onetypeofsensor, they areparticularlyusefulinseveral ing the GADGET-Matrix model. eTh hierarchical levels of the aspects of DBM. For instance, cameras focused on the driver GADGET-Matrix model consist of the Vehicle Maneuvering can be used to predict the driver’s state and fatigue levels. level mapped to the operational level in Michon’s model. It eTh y can also be used to improve maneuver recognition by mainly accounts for the drivers capability of operating the incorporating cues of the drivers eye gaze, hand position, and vehicle such as controlling of speed, the vehicle’s direction, foot hovering. Examples of such maneuvers include intent to andbraking.Themastering tracs ffi ituationslevel (mapped change lanes, brake, and turn that can be inferred earlier with to the tactical level) is mainly related to the drivers’ thinking the use of cameras as drivers check their blind spots and grip skills, which allow drivers to adapt to the current tracffi the steering wheel prior to taking action. situation. eTh thirdlevel is thegoals andcontext of driving level (mapped to the strategic level), which includes the tools 2.3. Modeling Levels. In driver behavior literature, several that evaluate the purpose and the environment of driving, models have been proposed; examples of these models are the that is, driving rules and where and when to drive. The hierarchical control model, the GADGET-Matrix model, and top level considers the importance of driving for the driver the DRIVABILITY model. eTh hierarchical control model is that motives and allows describing behaviors which are “less based on Michon’s theory. It has commonly been categorized congruent with the norms of the society” [21]. as operational, tactical, and strategic based on the timescales The DRIVABILITY model [22] is different from the through which they operate [2, 7]. aforementioned models by mainly focusing on the strategic model. The model describes driving behavior as a result to vfi e permanent and temporary contributors, which simulta- 2.3.1. Operational Level. Modeling operational maneuvers neously affect a driver’s decisions: involves actions performed over less than a second primarily in order to remain safe or abide by traffic regulations. Sudden Individual Resources. They are physical, social, psy- braking and turning are examples of this modeling domain, chological, and mental conditions of a driver. which operate at the shortest timescale of human interaction. Knowledge and Skills. They are the driver’s train- Such models canbeusedtoimprove vehicledesign, human- ing, education, experience, and knowledge not only vehicle interaction, and emergency assistance systems. An related to driving skills but in general, since these overview of such modeling techniques can be found in [15, factors greatly influence motivation and behavior of 16]. the driver. Environmental Factors. They include the vehicle sta- 2.3.2. Tactical Level. Tactical maneuvers can be defined as a tus, the existence of traffic hazards, the weather, and coherent set of operational maneuvers intended to achieve road and traffic conditions. a short-term goal such as lane changes, turns, and stops. eTh se operations typically last for several seconds, thereby Workload and Risk Awareness. eTh yare themaintwo enabling predictive modeling and inference. Modeling and key elements that tie the drivers’ resources to their predicting tactical maneuvers have significant potential to environmental status to facilitate understanding and improve ADASs since there is time to prevent unsafe driving analyzing driving performance. behaviors if the drivers are unaware of the danger of their actions. As such, models that enable early prediction of driver 2.4. Reactive and Predictive Models. DBM can be classified as intent prior to a tactical maneuver are of particular interest. either reactive or predictive models. Reactive models learn the An interesting survey of tactical maneuvers with emphasis on observed behavior or driving maneuver after the action has modeling driver intent can be found in [2]. been conducted. For instance, driver coaching applications can employ reactive models that identify dangerous driving 2.3.3. Strategic Level. At the strategic level, actions are trig- maneuvers performed by the trainee during the training gered by the long-term goals of the driver. For instance, session. On the other hand, predictive models are required destination and route calculation is an example of strategic to identify thedriveractiononthe onsetofthe behavior in actions where the timescale extends to minutes or hours real time. This is needed in ADAS where precautionary action [17–19]. Driver preferences can also be considered within should be performed immediately. eTh success of predictive 4 International Journal of Vehicular Technology models is contingent on how early they can predict the driver systems for novice drivers and to retrain elderly drivers by behavior,and they aretherefore typicallymoredicffi ult to understanding their deficiencies at different levels [8, 26]. develop than reactive models. 3.2. Driver Assistance Systems. As mentioned in the Intro- 2.5. Algorithms and Approaches. Algorithms and approaches duction, the majority of the driver fatalities and injuries are for DBM encompass a broad range of statistical, machine caused due to driver inattention and unintended maneuvers. learning, and pattern recognition techniques, among others. ADASs are thus being developed by industry and academic We highlight some of the most commonly used approaches projects in an effort to reduce or eliminate at best these below. casualties.Theprimary object of ADASsistoforecastthe trajectory andbehaviorofavehicleinrealtimeand then compensate for dangerous circumstances or events. To do 2.5.1. Basic Statistical Classification. Statistical models can be so,itisessential forthe ADASstobecapable of dieff renti- used to study the behavior of drivers based on collected data. ating between potentially dangerous situations and regular Simple trends in the data can be used to gain insight on driving behavior. Accurately modeling deceleration behavior the anticipated driver maneuvers and classification criteria is one element of such systems [27]. A primary challenge can be identified. Model tfi ting and regression techniques are however is to develop such systems without annoying the some common examples of such methods. While statistical driver with irrelevant recommendations and precautions or classification approaches are generally intuitive, they may be misinterpreting the state of the driver or the surrounding limited in their ability to classify complex multidimensional vehicles. Research in ADASs that involves multiple vehicles data [23]. canleadtomodelsthatcaptureright-of-wayrulesandgeneral road scene-awareness. Eventually Driver Assistance Systems 2.5.2. Discriminative Approaches. Discriminative approaches may evolve to driver-less systems for either semiautonomous such as Support Vector Machines (SVMs) are generally used or fully autonomous vehicles [28]. to overcome some of the limitations of basic classification schemes. SVMs can be used to ecffi iently model driver 3.3. Energy Efficiency. Driver behavior models can also behaviors where binary classification is involved such as be applied towards improving vehicle energy efficiency by determining driver compliance to tracffi rules or deciding monitoring the pedal actuation and fuel usage. Reports whether a driver will make a particular maneuver. Two and recommendations can then be provided to the driver. particular advantages of SVMs are as follows: (1)theysolve Additionally, optimizing electric vehicle sharing has been an optimization of a convex function, and thus the derived recently proposed in the literature [29]. solution is a global optimum, and (2)the upperbound on the generalization error does not depend on the problem dimension [23, 24]. 3.4. Crowdsourced Sensing for Road Conditions. Traditional research in DBM has focused on input from a single driver. eTh current direction of crowdsourced sensing and big data 2.5.3. Generative Models. Generative approaches are another analysis can be coupled with driver behavior models to gain primary modeling technique in DBM. Here, the underlying insight on the current road conditions. This includes tracffi patterns in the collected driver data are investigated and the jams,roadtypes,and speed limits,aswellaspredictingthe probability of observing a set of outputs for a given model weather conditions and degree of slipperiness [30]. is determined. Hidden Markov Models (HMMs) are one example where the relationship between the observations and the hidden states that generate these observations can be 4. Modeling Objectives identified [25]. Here, the states of the HMMs define different behaviors and the transitions between these states capture the While the applications discussed in the previous section evolution of the driver model. demonstrate the desired uses of DBM, they are typically achieved by individual modeling objectives which we review in this section. The objectives discussed herein are not meant 3. Applications to be comprehensive but rather representative of the major classes of DBM objectives. Modeling driver behaviors enables a plethora of applications facilitated by the constant advances in sensing and computa- tional capabilities. We discuss the recent developments and 4.1. Lane Changing. Lane changing models describe the applications in this section and summarize our discussion in drivers’ lane changing behaviors under various traffic con- Table 1. ditions. The primary goal is to determine whether or not it is safe for a driver to make a lane change given the 3.1. Driver Training and Self-Coaching. Many of the driver vehicle’s speed and the surrounding traffic. eTh gap acceptance models are developed aiming at facilitating better driver measure is a traditional approach used in lane changing training models. eTh idea is to monitor driver actions either models. A driver will only make a lane change if both the in a simulator or in a real environment and assess the driver lead and lag gaps in the target lane are above the safety safety andcompetencelevelsbased on models forideal threshold. eTh re are several challenges however that make driving. There has been particular interest in developing such lane changing models complicated such as the variance of International Journal of Vehicular Technology 5 ff Th fi fi fi Table 1: Summary of reviewed DBM approaches and challenges. Modeling Approaches Challenges and Directions (i) Lane changing literature reviews and classification [4, 5]. (ii) Rule-based approaches using Gipps Model with the lane changing process as a decision tree with a series of xed conditions [31]. Other (i) Incorporating personal driving incentives and preferences, with rule-based schemes include Cellular automata [32] and game theory based contextual factors such as weather and lighting, is needed to develop more models [33]. Lane changing personalized lane changing models. (iii) Discrete-choice models based on probabilities include [34, 35]. (ii) Works addressing more the less common and complex driving tasks (iv) Fuzzy-logic and artificial neural networks have been used in [4, 36, 37], such as ramp merging and multiple lane changing. to account for uncertainty and facilitate unsupervised training on real data. (v) Incentive-based models that incorporate factors such as the desire to follow a route, gain speed, and keep right [39] and politeness factors [38]. (i) Identifying the degree of stopping violations at intersections based on Speed Distance Regression (SDR) classification [52]. Field-test based characterization of driver stopping decisions for different age groups and (i) Developing cooperative models that leverage information from multiple genders [53]. vehicles to enable collective behavior at intersections. (ii) Driver behavior classification at intersections based on SVMs and (ii) Coupling intervehicular communications with driver behavior HMMs. Validation performed on a large naturalistic dataset [23]. modeling. Intersection decision-making (iii) Recognizing other behaviors at intersections, for example, turning and (iii) Developing unified standard datasets to evaluate different intersection stopping [42, 43] and left turns at signaled intersections [44]. behaviors. is will offer a platform for researchers to compare and (iv) Predicting multiple situations using case-based reasoning [45]. evaluate their modeling techniques. Modeling the evolution of an intersection using situation assessment and behavior prediction [46]. (i) Detecting aggressive driver behavior and competence using probabilistic ARX models [54]. (ii) Measuring in-vehicle acceleration using smartphone sensors to count (i) Trading o accuracy versus cost for use of traditional cell-phone sensors events of sudden acceleration, braking, and sharp turning [10]. versus advanced in-vehicle sensors and OBD. Driver profiling (iii) Fuzzy-logic based scoring mechanisms to prole driver aggressiveness (ii) Differentiating between aggressive behaviors and skilled maneuvers [11]. that include acceleration. (iv) Using an onboard diagnostic reader and an inertial measurement unit along with a Bayes classier to model aggressiveness [12]. (i) Survey on the current literature on route choice models with the focus on using fuzzy-logic and genetic algorithms [55]. (i) Building distributed end-to-end travel assistance systems that (ii) Classification of literature according to the considered user preferences, incorporate real-time sensing of traffic, weather, and road conditions. Router choice modeling for example, travel time, the number of intersections, traffic lights, and (ii) Developing more sophisticated personalized navigation and travel roadside aesthetics [56]. systems that learn and model user preferences. (iii) Incorporating feedback to learn user preferences [58] and exploring the evolution of driver route choices with time [57]. 6 International Journal of Vehicular Technology the gap acceptance behavior under different tracffi conditions the risk associated with the lane change [38]. Modeling and the dependence on the capabilities and types of the theincentivescan includeavarietyoffactors such as the vehicles. desire to follow a route, gain speed, and keep right [39], eTh re is a very large body of research on lane changing in addition to politeness factors [38] that can be tuned to models—a few literature surveys [4, 5] have been conducted account for different driver personalities. While incentive- to classify the different approaches. According to Rahman based approaches capture the human element of maximizing et al. [4] there are four categories of models: rule- personal benefits and driving preferences they lack more based models, discrete-choice probabilistic models, artificial- detailedphysicalmodelingthatmay limitapplicability to intelligence models, and incentive-based models. The Gipps different traffic situations such as congestion. Model [31] is among the most notable rule-based models. As discussed, several types of lane changing models It models thelanechangingprocess as adecisiontreewith have been proposed in the literature. However, novel models a series of xe fi d conditions that are typically found on the that combine the personal driving aspect (incentives and roadways and the output is a binary choice that indicates the preferences), road congestion, and geometric considerations, lane changing decision. Gipps Model incorporates a number as well as contextual factors such as weather and lighting, of logical and practical reasons into lane changing tree such are needed to develop generic lane changing models. Works as intent of turning, presence of heavy vehicles, existence of addressing specific lane changing maneuvers such as ramp thesafetygap,and speed advantageand hasbeenusedin merging [40, 41] and multiple lane changing are another open several microscopic tracffi simulation tools. However, Gipps area of research. Model does not include mechanisms to deal with different In addition to determining stopping, estimating the tracffi conditions. For instance, during congestion drivers general driver behavior at intersections is of significant may cooperate to allow other drivers to change lanes, or importance. Various works have addressed specific goals on the other hand drivers may be aggressive and force lane such as recognizing turning and stopping maneuvers [42, 43] change in theoretically an unsafe way. Cellular automata and left turns at signaled intersections [44]. In [45] Vacek [32] and game theory based models [33] have also been et al. consider the more challenging problem of predicting developed as rule-based models. The work in [33] tackles multiple situations using case-based reasoning. Modeling the cooperative and forced lane changes using game theory with evolution of an intersection situation is also investigated in driver experience as a parameter. [46] where the authors propose a multiple stage approach that Discrete-choice models basedonprobabilities have also combines situation assessment with behavior prediction. In received considerable attention in the literature. Ahmed the rfi st stage, the current intersection situation is classiefi d [34] developed a generic lane changing model that captures by decomposing it into more manageable sets of related road both mandatory and discretionary lane changes in a simple users to prevent a combinatorial explosion of variables. eTh mathematical formulation. Toledo et al. [35] also propose a interactions between the entities are used to determine the probabilistic model where the trade-offs between forced and configuration; for example, a vehicle has to slow down to discretionary maneuvers are combined in a single utility and keep asafedistanceorstopatatracl ffi ight.Subsequently, tuned using maximum-likelihood estimation approaches. in the second stage the velocity profile of each vehicle Extensive tests on microscopic vehicle trajectory data col- is predicted taking advantage of the previously estimated lected in Arlington, USA, have confirmed the effectiveness of situation using random forest regressors [47]. More recently, Toledo’s models. a two-layer framework for estimating driver decisions at Several lane changing models based on fuzzy-logic and intersections has been proposed [48]. As opposed to the top- articfi ial neural networks have also been developed, although down approach of estimating the intersection situation and their adoption remains limited [4]. eTh advantage of fuzzy- then using the underlying continuous model to determine logic is that the uncertainty in lane changing can be modeled thevehicle dynamics (asin[46]),the authorspropose and a number of abstract IF-THEN rules can be used to bottom-up architecture. Their reasoning is that it is easier represent the complex decision-making. Among the more to observe the lower level states such as vehicle position, recent works is that of Moridpour et al. [36] which focuses velocity, orientation, and yaw. These continuous observations on lane change behavior of heavy vehicles. Neural networks are modeled as Gaussian Mixture Models (GMMs), and the have demonstrated high accuracy in modeling lane changing higher level discrete state system is modeled using HMMs decisions on eld fi collected data [37]. Inputs such as the vehi- corresponding to the potential driver decisions. cle’s direction, speed, distance from surrounding vehicles, and preferred speed have been used to train the network 4.2. Intersection Decision-Making. Reports indicate that an using the backpropagation algorithm with promising results. estimated 45% of injury crashes and 22% of roadway fatalities The primary disadvantage of artificial-intelligence based were intersection related in the United States [49]. Such statis- approaches is the dependence on field collected data for tics have driven several international research projects that different tracffi situations in order to calibrate and develop specifically target intersection decision collision avoidance the models satisfactorily. systems [44, 50, 51]. Incentive-based models have been more recently consid- A primary objective in signalized intersection decision- ered in modeling lane changing behavior. In essence, the making is to predict whether the driver will stop safely incentive criterion models the attractiveness of a lane based before the stop bar if the signal turns red. This classification on its utility to the driver, and a safety criterion captures is then integrated into ADASs to warn drivers of the own International Journal of Vehicular Technology 7 violations and may also be used to warn other drivers via basis. These services typically provide one or more alternative V2V and Vehicle-to-Infrastructure (V2I) communication. routes primarily based on the shortest distance between the eTh observations typically needed are the vehicles position, source and destination. However, in reality there are several speed, and acceleration and the tracffi signal phase that other factors that inu fl ence the preferred route for a user. For are monitored over a time window [23]. Time Transmis- instance, different drivers may have varying comfort levels sion Interval (TTI), Time-to-Collision (TTC), and Required with driving along highways, making multiple lane changes, Deceleration Parameter (RDP) are also commonly used in or leftturns at tracl ffi ights. Thisisimportant fornovice intersection safety systems. The vehicles TTI =𝑟/ V,where drivers and the older population who are in particular need of V is the vehicle’s current speed and𝑟 is its distance to the using mapping services. In addition to the driver competence intersection. eTh TTI is computed on the onset of braking level, personal preferences can also play a significant role and compared to the required time for a safe stop. In a in route selection. This includes the number of controlled similar approach, RDP denotes the required deceleration for intersections, stop signs, or routes with frequent public the vehicle to stop safely given its current distance and speed transportation stops and school buses. u Th s, while mapping and a comparison is made to RDP threshold to determine services have revolutionized the user navigation experience if therequireddecelerationislargerthanthatpermissible. today, much research and development are needed towards Aslightlymoreinvolvedclassicfi ation basedonSDR has personalized travel information and Driver Assistance Sys- been presented in [52]. Here, instead of making independent tems. observations, a regression curve is generated from a set A recent survey of the current literature on route choice of speed and distance measurements of compliant vehicles. models in transportation networks has been covered in [55]. Measurements are then compared to the compliant SDR According to the survey, several fuzzy-logic and reasoning curves to identify the degree of violations and issue warnings. approaches have been adopted due to their simplicity in In [53] field tests are conducted to characterize the driver dealing with uncertainty and qualitative variables. Genetic stopping decisions for different age groups and genders. More algorithms and ant colonization were also considered in sev- sophisticated models based on SVMs and HMMs have also eral applications of route-finding problems in transportation been developed for higher classification accuracy [23]. networks. Another interesting literature review is available In summary, modeling driver behavior at intersections is in [56] where Ramaekers et al. classify the works according indeed a very complex task, and a lot of further research is to the factors considered in the models such as travel time, needed before the goal of fully autonomous driving can be number of intersections, tracffi lights, roadside aesthetics, achieved. and several other factors. This study also investigates the relationship between the purpose of the trip and the road categories used. A major limitation of previous studies is the 4.3. Driver Prolfi ing and Characterization. The objective of assumptions of perfect knowledge due to lack of information driver profiling is not to model specific maneuvers but aboutthe transportnetwork.Thus,incorporating real-time general driver characterization. Examples include detecting information from a cloud-sourced sensing platform is one aggressive driver behavior [54] and the level of driver compe- open area of research. Learning user preferences is another tence by assessing behavior in difficult situations/maneuvers relevant area of research where Tawfik et al. [57] explore such as driving on ice and avoiding accidents. Driver profiling the evolution of driver route choices with time. eTh authors can be integrated into ADASs for vehicles with multiple conclude that while some drivers maintain their choice, drivers to tailor the assistance recommendations to each others arekeentocontinuouslyevaluatealternative routes.A driver. detailed analysis is made where factors such as ethnicity, edu- A large body of literature is available on driving behav- cation, driving experience, and gender are included. These ior analysis to determine aggressive behaviors and provide results dictate that personalized travel information systems safety recommendationsinDriverAssistanceSystems [10– are needed to cope with such dieff rences. In such systems, 12, 14]. For instance, the work in [10] measures in-vehicle collecting user feedback is elemental and is concluded in the acceleration using smartphone sensors to count events of work of Park et al. [58]. sudden acceleration, braking, and sharp turning. eTh authors emphasize the need for dynamic calibration algorithms when using phone sensors. More recent efforts have also been con- 5. Evaluation Methodologies ducted in [11] where a fuzzy-logic based scoring mechanism is introduced to profile driver aggressiveness on a scale of In this section we discuss the common evaluation techniques [0,100]. In addition to using smartphone sensors, an onboard of DBM that include real datasets and driving simulators, as diagnostic reader and an inertial measurement unit were well as some of the typical metrics. used in [12] along with a Bayes classifier to model aggressive driving behaviors. 5.1. Microscopic Datasets. Datasets of driver behavior are typically generated by collecting vehicle and sensor data of 4.4. Route Choice Prolfi ing and Travel Assistance. Navigation multiple subjects as they drive. eTh collected datasets are systems and online maps have garnered increasing user further classified as naturalistic and instructed. In naturalistic adoption over recent years, with statistics indicating that 55% data collected, the subjects are told to drive as they normally of smartphone holders use mapping services on a regular wouldwhere only theroute maybespeciefi dbut thegoals of 8 International Journal of Vehicular Technology thestudy arenot.This enablesthe most natural form of data current literature. This includes but is not limited to for use in ADASs. On the other hand, instructed data col- ramp merging, multiple lane changing, cooperative lection typically involves informing the subjects to perform intersection behavior, and driver intention modeling. specicfi driving tasks or scenarios and monitoring the output. (iii) Although there are several available datasets for Thisallowsmorefocus andemphasisoncollectingdataof DBM evaluation, more work is needed towards a particular maneuvers but potentially modiefi s the driving unified standard dataset for different applications. behavior since the subjects are consciously repeating certain This will offer a platform for researchers to compare tasks. Specific datasets are typically created for different andevaluatetheir modeling techniques.Thus,more driving tasks/objectives. For instance, references to datasets representative data from field tests for drivers of for driver behavior at intersections can be found in [50, 59], different genders and age groups is without a doubt while several lane changing datasets are discussed in [4]. also required. (iv) Personalized navigation and travel systems that learn 5.2. Simulators. Several DBM studies have been based on and model user preferences are another challenging simulator studies as well. eTh advantages of driving simu- area of research. Previous works in this direction lators are that several variables such as distractions, tracffi have primarily assumed perfect knowledge of the conditions, and weather can be controlled accurately without road network and environment, which is not realistic. compromising safety. Simulator-based studies are also eas- eTh refore, incorporating real-time information from ily repeatable and facilitate large data collection. However, a cloud-sourced sensing platform will foster greater simulator-based models may not accurately reflect the on- readiness for practical implementation. road performance and therefore validation with real data is (v) Personalized driver monitoring and state recogni- needed. tion that can capture drivers’ state dynamically and online. Previous proposals make use of vehicle or/and 5.3. Metrics. True and false positive rates are common met- smartphone sensors and the driver profile to detect rics for evaluation in DBM where a specific task is to be drivers’ abnormal states such as drowsiness. Most predicted. True positive rates represent the percentage of drowsiness detection schemes assume that the driver correctly predicted events while false positive rates denote is always facing the camera or ignore the level of the percentage of events that were incorrectly predicted illumination, which directly aeff cts the correctness as true. While these statistics provide an overview of the of the collected data for image processing, which model’s success, it is important to consider the details of the render these proposals impractical. Other methods testing and learning environment of each study in order to such as context-aware schemes could be explored objectively compare performances objectively. eTh timeliness to recognize the driver state by detecting abnormal of the prediction is another important parameter used in actions such as zigzag pattern driving and random DBM that indicates the proactive capabilities of the different and risky acceleration and lane changes. In addition, models. Naturally, as the time gets closer to the maneuver, the simpler and dynamic image processing methods are performance increases. required to detect the driver’s states online. 6. Open Research Challenges and 6.2. Emerging Directions Future Directions 6.2.1. Cooperative Modeling Approaches for Collective Scene Driver behavior modeling is currently receiving increasing Modeling and Sensing. Most of the current literature on interest from industry and academia due to several contem- driver behavior modeling has focused on a single vehicle porary factors, including the challenges of increasing global making inferences based on sensed measurement of the urbanization and demand for smart infrastructure solutions, driver, the vehicle, and its environment. Today, advances the emergence of enabling technologies such as advanced in vehicle-to-vehicle (V2V) and Vehicle-to-Infrastructure sensing and data analytics, and the demand for futuristic (V2I) communications can facilitate novel approaches to applications such as autonomous vehicles. In this section we driver behavior modeling. In particular, cooperative models first summarize some of the current open research challenges can be developed that leverage information from multiple and then highlight two emerging directions of future research vehicles to develop more global behavioral models. This can in Driver Assistance Systems (DASs). enable driving-scenario or situation modeling for diverse applications and scenarios such as collective behavior at 6.1. Research Challenges intersections [60]. For instance, coupling intervehicular com- munications with driver behavior modeling can facilitate the (i) Novel DBM that incorporates personal driving incen- following advancements in DBM. tives and preferences, with contextual factors such as weather and lighting, is needed to develop more Signaling Warnings. In addition to alerting the driver to dan- personalized and generic models. gerous situations, these can be communicated to surrounding (ii) Works addressing more the less common and com- vehicles.Previouseoff rtsinADAShaveprimarily focusedon plex driving tasks are needed to complement the alerting the driver and not the surrounding vehicles. Some International Journal of Vehicular Technology 9 Road network modeling Driver Assistance Semistatic road Road network Cloud (DAC) information representation Criteria selection Crowdsourced and modeling sensing platform Dynamic environmental information Driver Driver services center profiling unit Learning Route selection route Communicating algorithms preferences preferences and driver feedback Determining Driver alerts and driver skills special services Speed, vibrations, acceleration, etc. Figure 2: Overview of a personalized Driver Assistance Cloud (DAC). recent works that investigate the use of V2V communication needs, and it is necessary to identify the beneficial to signal warnings include [23, 61]. information exchange. (ii) How can multiagent models be developed that aggre- Scene Modeling by Information Exchange.Inessence,road gate the information from multiple sources? As single driving is a collaborative action. eTh behavior of one driver driver behavior modeling is already a difficult task will impact the behavior of others. u Th s, building models that involves several parameters, simple multidriver that simultaneously incorporate inputs from multiple drivers modeling approaches should be developed rfi st. can generate a model that can more accurately predict the collective behavior. With intervehicular communications (iii) How can feedback be effectively incorporated to such models can be derived, and we refer to this as scene or model the scene evolution as time progresses? road situation modeling. 6.2.2. Personalized Driver Assistance Clouds. While several Early Model Building. eTh sensors used to model driver developments have independently been made in features for behavior for Driver Assistance Systems typically have a in-vehicle ADAS, there is limited work towards a framework range of a couple of hundred meters. Vehicle-to-vehicle that integrates the current sensing capabilities, driver behav- communication can expand the sensing range further and ior models, and communication to the cloud. Such a system, enable models that can predict driver behavior early on. which we refer to as a Driver Assistance Cloud, can provide Collaborative Objectives. Applications such as adaptive cruise novel personalized driver services, applications, and safety, as illustrated in Figure 2. control can benefit significantly from intervehicular commu- In order to develop such systems a road traffic infor- nications where long-term planning based on the positions of several vehicles can be beneficial. In such cases, fuel efficiency mation repository can be created to integrate environmental and road attributes such as weather conditions, construction, can be improved by optimizing the speed over a time horizon. In order to develop cooperative driver behavior models prevalence of pedestrians, bus stops, and potholes. These attributes can be incorporated from diverse sensing sources we need to answer the following research questions: aer ft intricate calibration and pruning. DBM that provides (i) What information is relevant for intervehicular com- insight on the driver skill level for different maneuvers will munication? Different applications will have different then be needed. For instance, acceleration profiles can be 10 International Journal of Vehicular Technology analyzed to determine lane changing and ramp merging opportunities,” IEEE Transactions on Intelligent Transportation Systems,vol.14, no.4,pp. 1942–1956, 2013. competence. A particular feature of DACs may be designing efficient route selection algorithms based on driver profiles. [5] S.Moridpour,M.Sarvi,and G. Rose,“Lane changing models:a This includes rfi st learning route preferences based on moni- critical review,” Transportation Letters,vol.2,no. 3, pp.157–173, toring both the routes taken by drivers and their competence levels on different road types. eTh n, multicriteria decision- [6] W. Wang, J. Xi, and H. Chen, “Modeling and recognizing making techniques can be applied to determine the routes driver behavior based on driving data: a survey,” Mathematical Problems in Engineering,vol.2014, ArticleID245641, 20 pages, most suited to different drivers. [7] T. A. Ranney, “Models of driving behavior: a review of their 6.2.3. DBM for Level 3 Automated Driving. Level 3 automated evolution,” Accident Analysis & Prevention,vol.26, no.6,pp. driving [62] refers to vehicles that are automatically equipped 733–750, 1994. to control all driving functions with little to no attention [8] M. Panou, E. Bekiaris, and V. Papakostopoulos, “Modelling of the driver for specicfi periods. In level 3 automated driver behaviour in european union and international projects,” driving human intervention is expected at any moment at the in Modelling Driver Behaviour in Automotive Environments,pp. human’sdiscretion. Thisisanewareaofresearchwithseveral 3–25, Springer, Berlin, Germany, 2007. nontrivial DBM challenges: [9] I. Hawkins, Torque Pro (OBD 2 & Car), https://play.google .com/store/apps/details?id=org.prowl.torque&hl=en. (i) Rapid onboarding: modeling of the driver behavior of reestablishing the driving context when switching [10] J. Paefgen, F. Kehr, Y. Zhai, and F. Michahelles, “Driving behavior analysis with smartphones: insights from a controlled from automated driving to human driving field study,” in Proceedings of the 11th International Conference (ii) Complexity: the fact that fully automated vehicle on Mobile and Ubiquitous Multimedia (MUM ’12),p.36, ACM, results in increased complexity of the vehicle func- Ulm, Germany, December 2012. tionality and communication. Fully automated vehi- [11] G. Castignani, R. Frank, and T. Engel, “An evaluation study cles are expected to utilize VANET communication, of driver profiling fuzzy algorithms using smartphones,” in which is too fast for human to monitor. Consequently, Proceedings of the 2nd International Workshop on Vehicular careful DBM design and analysis are needed espe- Communications and Applications (VCA ’13),p.6,Goettingen, cially on the event of automated system error when Germany, October 2013. the human intervention is needed in short time. [12] J.-H. Hong, B. Margines, and A. K. Dey, “A smartphone-based sensing platform to model aggressive driving behaviors,” in (iii) Cooperative DBM between level 3 automated driving Proceedings of the 32nd Annual ACM Conference on Human and levels 2 and 1 automated driving Factors in Computing Systems (CHI ’14), pp. 4047–4056, May Competing Interests [13] R. Arau´jo, A. Igreja, R. de Castro, and R. E. Araujo, “Driving coach: a smartphone application to evaluate driving efficient eTh authors declare that there is no conflict of interests patterns,” in Proceedings of the Intelligent Vehicles Symposium regarding the publication of this paper. (IV ’12), pp. 1005–1010, Alcala´ de Henares, Spain, 2012. [14] C.-W. You, N. D. Lane, F. Chen et al., “CarSafe App: alerting drowsy and distracted drivers using dual cameras on smart- Acknowledgments phones,” in Proceedings of the 11th Annual International Confer- This work is funded by Project no. 31R014-Research Center- ence on Mobile Systems, Applications, and Services (MobiSys ’13), RTTSRC-4-2013providedbythe RoadwayTransportation pp. 13–26, ACM, June 2013. & Traffic Safety Research Center, United Arab Emirates [15] M. Ploc ¨ hl and J. Edelmann, “Driver models in automobile University. dynamics application,” User Modeling and User-Adapted Inter- action,vol.45, no.7-8,pp. 699–741, 2007. [16] C. C. Macadam, “Understanding and modeling the human References driver,” Vehicle System Dynamics,vol.40, no.1–3,pp. 101–134, [1] United Nations, “World’s population increasingly urban with more than half living in urban areas,” http://www [17] J. Froehlich and J. Krumm, Route Prediction from Trip Observa- .un.org/en/development/desa/news/population/world-urbani- tions, Society of Automotive Engineers (SAE) World Congress, zation-prospects-2014.html. [2] A. Doshi and M. M. Trivedi, “Tactical driver behavior predic- [18] R. Simmons, B. Browning, Y. Zhang, and V. Sadekar, “Learning tion and intent inference: a review,” in Proceedings of the 14th to predict driver route and destination intent,” in Proceedings of IEEE International Intelligent Transportation Systems Confer- IEEE Intelligent Transportation Systems Conference (ITSC ’06), ence (ITSC ’11), pp. 1892–1897, October 2011. pp.127–132,Toronto,Canada, 2006. [3] L. S. Angell, J. Auflick, P. Austria et al., “Driver workload metrics [19] S. Peeta and J. W. Yu, “Adaptability of a hybrid route choice task 2 final report,” Tech. Rep. DOT HS 810635, NHTSA, US model to incorporating driver behavior dynamics under infor- Department of Transportation, 2006. mation provision,” IEEE Transactions on Systems, Man, and [4] M. Rahman, M. Chowdhury, Y. Xie, and Y. He, “Review Cybernetics Part A:Systems and Humans.,vol.34, no.2,pp. 243– of microscopic lane-changing models and future research 256, 2004. International Journal of Vehicular Technology 11 [20] M. Hatakka, E. Keskinen, E. Katila, and S. Laapotti, “Do [37] A. Dumbuya, A. Booth, N. Reed et al., “Complexity of traffic psychologists have something to oeff r in driver training, driver interactions: improving behavioural intelligence in driving sim- improvement and selection?” 1997. ulation scenarios,” in Complex Systems and Self-Organization Modelling, pp. 201–209, Springer, Berlin, Germany, 2009. [21] S. Laapotti, E. Keskinen, M. Hatakka et al., “Driving circum- stances and accidents among novice drivers,” Tracffi Injury [38] A. Kesting, M. Treiber, and D. Helbing, “General lane- Prevention,vol.7,no. 3, pp.232–237,2006. changing model MOBIL for car-following models,” Transporta- tion Research Record: Journal of the Transportation Research [22] E. Bekiaris, A. Amditis, and M. Panou, “Drivability: a new con- Board,vol.1999,no. 1, pp.86–94,2007. cept for modelling driving performance,” Cognition, Technology &Work,vol.5,no. 2, pp.152–161,2003. [39] W. Schakel, V. Knoop, and B. Van Arem, “Integrated lane change model with relaxation and synchronization,” Transportation [23] G. S. Aoude, V. R. Desaraju, L. H. Stephens, and J. P. How, Research Record, vol. 2316, pp. 47–57, 2012. “Driver behavior classification at intersections and validation on largenaturalisticdataset,” IEEE Transactions on Intelligent [40] S. Bonnin, T. H. Weisswange, F. Kummert, and J. Schmuud- Transportation Systems,vol.13, no.2,pp. 724–736, 2012. derich, “Accurate behavior prediction on highways based on a systematic combination of classifiers,” in Proceedings of the IEEE [24] V. Vapnik, The Nature of Statistical Learning eTh ory ,Springer, Intelligent Vehicles Symposium (IV ’13), pp. 242–249, June 2013. Berlin, Germany, 2000. [41] Y. Hou, P. Edara, and C. Sun, “Modeling mandatory lane chang- [25] L. R. Rabiner, “A tutorial on hidden markov models and selected ing using bayes classifier and decision trees,” IEEE Transactions applications in speech recognition,” Proceedings of the IEEE,vol. on Intelligent Transportation Systems,vol.15, no.2,pp. 647–655, 77,no. 2, pp.257–286,1989. [26] K. Takeda, C. Miyajima, T. Suzuki et al., “Self-coaching system [42] E. Kaf ¨ er, C. Hermes, C. Woh ¨ ler, H. Ritter, and F. Kummert, based on recorded driving data: learning from one’s experi- “Recognition of situation classes at road intersections,” in ences,” IEEE Transactions on Intelligent Transportation Systems, Proceedings of the IEEE International Conference on Robotics and vol. 13,no. 4, pp.1821–1831,2012. Automation (ICRA ’10),pp. 3960–3965, May2010. [27] P. Angkititrakul, C. Miyajima, and K. Takeda, “Analysis and [43] M. Huls ¨ en, J. M. Zol ¨ lner, and C. Weiss, “Traffic intersection prediction of deceleration behavior during car following using situation description ontology for advanced driver assistance,” stochastic driverbehavior model,” in Proceedings of the 15th in Proceedings of the IEEE Intelligent Vehicles Symposium (IV ’11), International IEEE Conference on Intelligent Transportation Sys- pp. 993–999, IEEE, Baden-Baden, Germany, June 2011. tems (ITSC ’12), pp. 1221–1226, Anchorage, Ala, USA, September [44] B. Bougler, D. Cody, and C. Nowakowski, “California inter- section decision support: a driver-centered approach to left- [28] A. Gray,Y.Gao,J.K.Hedrick,and F. Borrelli,“Robust predictive turn collision avoidance system design,” in Proceedings of the control for semi-autonomous vehicles with an uncertain driver California Partners for Advanced Transit and Highways (PATH model,” in Proceedings of the IEEE Intelligent Vehicles Sympo- ’08),January 2008. sium (IV ’13), pp. 208–213, Gold Coast, Australia, June 2013. [45] S. Vacek, T. Gindele, J. Zollner, and R. Dillmann, “Situation [29] S. D. Cairano, D. Bernardini, A. Bemporad, and I. V. Kol- Classification for Cognitive Automobiles using Case-based manovsky, “Stochastic MPC with learning for driver-predictive Reasoning,” in Proceedings of the IEEE Intelligent Vehicle Sym- vehicle control and its application to HEV energy management,” posium (IV ’07), pp. 704–709, Istanbul, Turkey, 2007. IEEE Transactions on Control Systems Technology,vol.22, no.3, pp. 1018–1031, 2014. [46] M. Platho, H.-M. Groß, and J. Eggert, “Predicting velocity [30] J. Straub, S. Zheng, and J. W. Fisher, “Bayesian nonparametric profiles of road users at intersections using configurations,” in Proceedings of the IEEE Intelligent Vehicles Symposium (IV ’13), modeling of driver behavior,” in Proceedings of the Intelligent Vehicles Symposium Proceedings, pp. 932–938, Dearborn, Mich, pp. 945–951, IEEE, Queensland, Australia, June 2013. USA, 2014. [47] L. Breiman, “Random forests,” Machine Learning,vol.45, no.1, [31] P. G. Gipps, “A model for the structure of lane-changing pp. 5–32, 2001. decisions,” Transportation Research Part B: Methodological,vol. [48] V. Gadepally, A. Krishnamurthy, and U. Ozguner, “A framework 20,no. 5, pp.403–414,1986. for estimating driver decisions near intersections,” IEEE Trans- actions on Intelligent Transportation Systems,vol.15, no.2,pp. [32] K. Nagel, D. E. Wolf,P.Wagner, andP.Simon,“Two-lane traffic rules for cellular automata: a systematic approach,” Physical 637–646, 2014. Review E, vol. 58, no. 2, pp. 1425–1437, 1998. [49] National Highway Traffic Safety Administration (NHTSA), [33] Y. Pei and H. Xu, “The control mechanism of lane changing “Fatality analysis reporting system encyclopedia,” vol. 17, 2010, in jam condition,” in Proceedings of the 6th World Congress on https://www-fars.nhtsa.dot.gov/. Intelligent Control and Automation (WCICA ’06),vol.2,pp. [50] M. Maile, F. Zaid, L. Caminiti, L. Lundberg, and P. Mudalige, 8655–8658, IEEE, Dalian, China, June 2006. Cooperative Intersection Collision Avoidance System Limited [34] K. I. Ahmed, Modeling drivers’ acceleration and lane changing to Stop Sign and Tracffi Signal Violations ,Phase 1, National behavior [Ph.D. thesis], Massachusetts Institute of Technology, Highway Traffic Safety Administration, Washington, DC, USA, Cambridge, Mass, USA, 1999. 2008. [35] T.Toledo,H.N.Koutsopoulos,andM.E.Ben-Akiva,“Modeling [51] K. Fuerstenberg, J. Chen, and S. Deutschle, “New European integrated lane-changing behavior,” Transportation Research approach for intersection safety—results of the EC-project Record, no. 1857, pp. 30–38, 2003. intersafe,” in Advanced Microsystems for Automotive Applica- tions 2007, VDI-Buch, pp. 61–74, Springer, Berlin, Germany, [36] S. Moridpour, M. Sarvi, G. Rose, and E. Mazloumi, “Lane- changing decision model for heavy vehicle drivers,” Journal of Intelligent Transportation Systems: Technology, Planning, and [52] H. Berndt, S. Wender, and K. Dietmayer, “Driver braking Operations,vol.16, no.1,pp. 24–35, 2012. behavior during intersection approaches and implications for 12 International Journal of Vehicular Technology warning strategies for driver assistant systems,” in Proceedings of the IEEE Intelligent Vehicles Symposium (IV ’07),pp. 245–251, Istanbul, Turkey, June 2007. [53] H. Rakha, I. El-Shawarby, and J. R. Setti, “Characterizing driver behavior on signalized intersection approaches at the onset of a yellow-phase trigger,” IEEE Transactions on Intelligent Transportation Systems,vol.8,no. 4, pp.630–640,2007. [54] M. Sundbom, P. Falcone, and J. Sjoberg, “Online driver behavior classification using probabilistic ARX models,” in Proceedings of the 16th International IEEE Conference on Intelligent Transporta- tion Systems: Intelligent Transportation Systems for All Modes (ITSC ’13), pp. 1107–1112, The Hague, The Netherlands, October [55] M. Sharma,J.K.Gupta,and A. Lala,“Survey of routechoice models in transportation networks,” in Intelligent Computing, Networking, and Informatics,pp. 1285–1290, 2014. [56] K. Ramaekers, S. Reumers, G. Wets, and M. Cools, “Modelling route choice decisions of car travellers using combined GPS and diary data,” Networks and Spatial Economics,vol.13, no.3,pp. 351–372, 2013. [57] A. M. Tawfik, H. A. Rakha, and S. D. Miller, “An experimental exploration of route choice: identifying drivers choices and choice patterns, and capturing network evolution,” in Proceed- ings of the 13th International IEEE Conference on Intelligent Transportation Systems (ITSC ’10), pp. 1005–1012, Madeira Island, Portugal, September 2010. [58] K. Park, M. Bell, I. Kaparias, and K. Bogenberger, “Learning user preferences of route choice behaviour for adaptive route guidance,” IET Intelligent Transport Systems,vol.1,no. 2, pp. 159–166, 2007. [59] T. Streubel and K. H. Hoffmann, “Prediction of driver intended path at intersections,” in Proceedings of the 25th IEEE Intelligent Vehicles Symposium (IV ’14), pp. 134–139, IEEE, Dearborn, Mich, USA, June 2014. [60] T. Gindele, S. Brechtel, and R. Dillmann, “A probabilistic model for estimating driver behaviors and vehicle trajectories in traffic environments,” in Proceedings of the 13th International IEEE Conference on Intelligent Transportation Systems (ITSC ’10),pp. 1625–1631, IEEE, Funchal, Portugal, September 2010. [61] S. Al-Sultan, A. H. Al-Bayatti, and H. Zedan, “Context-aware driver behavior detection system in intelligent transportation systems,” IEEE Transactions on Vehicular Technology,vol.62,no. 9, pp. 4264–4275, 2013. [62] S. M. Casner,E.L.Hutchins, andD.Norman, “ec Th hallengesof partially automated driving,” Communications of the ACM,vol. 59,no. 5, pp.70–77,2016. International Journal of Rotating Machinery International Journal of Journal of The Scientific Journal of Distributed Engineering World Journal Sensors Sensor Networks Hindawi Publishing Corporation Hindawi Publishing Corporation Hindawi Publishing Corporation Hindawi Publishing Corporation Hindawi Publishing Corporation http://www.hindawi.com http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 Volume 2014 Journal of Control Science and Engineering Advances in Civil Engineering Hindawi Publishing Corporation Hindawi Publishing Corporation http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 Submit your manuscripts at http://www.hindawi.com Journal of Journal of Electrical and Computer Robotics Engineering Hindawi Publishing Corporation Hindawi Publishing Corporation http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 VLSI Design Advances in OptoElectronics International Journal of Modelling & Aerospace International Journal of Simulation Navigation and in Engineering Engineering Observation Hindawi Publishing Corporation Hindawi Publishing Corporation Hindawi Publishing Corporation Hindawi Publishing Corporation Volume 2014 http://www.hindawi.com Volume 2014 Hindawi Publishing Corporation http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com http://www.hindawi.com Volume 2014 International Journal of Active and Passive International Journal of Antennas and Advances in Chemical Engineering Propagation Electronic Components Shock and Vibration Acoustics and Vibration Hindawi Publishing Corporation Hindawi Publishing Corporation Hindawi Publishing Corporation Hindawi Publishing Corporation Hindawi Publishing Corporation http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014
Journal
International Journal of Vehicular Technology – Unpaywall
Published: Dec 28, 2016