A Joint Trajectory and Signal Optimization Model for Connected Automated Vehicles
Amir Ghiasi, University of South FloridaShow Abstract
Xiaopeng (Shaw) Li, University of South Florida
Zhitong Huang, Leidos, Inc.
Xiaobo Qu, Chalmers University of Technology
Traffic conflict points are among the most common types of bottlenecks that cause travel delay, stop-and-go traffic, and excessive energy consumption. Connected automated vehicle (CAV) technologies make efficient control possible, and thus can be a cure to these issues. This paper formulates a joint vehicle trajectory and signal timing optimization model for guiding movements of CAVs on a general two-way conflicting points including signalized intersections and work-zones. This model simultaneously optimizes the CAV trajectories and signal timing to minimize travel delay and energy consumption. This study modifies the original complex problem in two ways. First, this paper simplifies the vehicle trajectory shape with a piece-wise quadratic function with no more than five segments. Second, instead of using the highly non-linear instantaneous fuel consumption function, this study proposes a simplified macroscopic measure that approximates fuel consumption as a function of signal red interval. These modifications provide elegant theoretical properties that enable an analytical exact solution. Numerical examples reveal that the proposed model can significantly reduce vehicle’s delay and fuel consumption. Moreover, it is demonstrated that the presented algorithm is highly efficient and appropriate for real-world traffic applications.
Multi-Class Traffic Flow Model Based on 3D Flow-Concentration Surface
Ranju Mohan, Indian Institute of ScienceShow Abstract
Gitakrishnan Ramadurai, Indian Institute of Technology, Madras
The paper proposes a continuum model for multi-class traffic based on a three dimensional flow-concentration surface. The propagation speed of small disturbance (PSSD) is derived based on the flow-concentration surface. Using the proposed PSSD and with a recently developed speed-area occupancy curve, a macroscopic continuum model for multi-class traffic is formulated that can replicate traffic flow behaviour where lane discipline is not strictly followed. The qualitative and quantitative properties of the proposed model are checked using numerical simulations for hypothetical road sections. The proposed model correctly reproduce the congestion formation and dissipation behaviour of multi-class traffic and the phenomena of local cluster effect.
Urban Road Network Performance with Shared Automated Vehicles
Henrik Becker, ETHZ - Swiss Federal Institute of TechnologyShow Abstract
Allister Loder, ETH Zurich
Kay W. Axhausen, Institute for Transport Planning and Systems
Automated vehicles are widely expected to bring about various benefits for (urban) transportation, e.g. by increasing road capacities and reducing generalized cost of travel. However, the latter may induce additional demand for road transportation, possibly counteracting gains in accessibility. Hence, the net impact of vehicle automation on road network performance is still unclear. This research uses the macroscopic fundamental diagram (MFD) to address this question for different levels of road capacity increases and modal splits between private and shared (i.e. public) transportation. To this end, various scenarios are tested in a simulation model for the morning peak hour for Zurich, Switzerland, as a case study, for which current demand levels for car and public transport are used. Yet, the results can be generalized to different city types. The analysis indicates that in car-oriented cities, vehicle automation will likely bring substantial benefits in network performance in current car-oriented cities, while for public transport oriented cities, substantial gains in road capacity of 40% or more will be required to make up for the potentially substantial mode shift from public transport towards (pooled) cars. Moreover, results show that up to 75% mode share of ride-sharing trips will be required to achieve a system-optimal state, which - depending on the technologically possible capacity gains - may even be inferior to the current state.
Implementation and Investigation of a Weather and Jam Density-Tuned Network Perimeter Controller
Maha Elouni, Virginia Polytechnic Institute and State UniversityShow Abstract
Hesham Rakha, Virginia Polytechnic Institute and State University
Youssef Bichiou, Virginia Polytechnic Institute and State University
This paper implements and evaluates the performance of a network fundamental diagram (NFD) proportional-integral perimeter controller (PC) using base tuned parameters (clear weather conditions and a base jam density of 160 veh/km/lane). The parameters were then re-tuned separately for different weather conditions and jam density values (reflecting different percentage of trucks), resulting in two new control methods: a weather-tuned perimeter controller (WTPC) and a jam density-tuned perimeter controller (JTPC). The WTPC was shown to outperform the no control strategy and the PC for different weather conditions. Specifically, the WTPC decreased the congestion inside the protected network (PN) and improved the overall performance of the full network (FN) by decreasing the average vehicle travel time, decreasing the vehicle total delay, increasing the average vehicle speed and decreasing the average vehicle fuel consumption. Alternatively, the JTPC was shown to perform similar to the PC for jam densities higher than the base-case jam density used in the PC (in our case 160 veh/km/lane). However, the JTPC outperformed the PC for smaller jam densities given that these smaller jam densities result in queues spilling back faster to upstream traffic signals. The results demonstrate the need to tune the controller to the actual jam density especially when the jam density decreases (e.g. trucks are introduced to the network).
Development of an Adaptive Automated Controller for Platoon Merging Using a Simulation-Based Approach
Sogand Karbalaieali, Fehr and PeersShow Abstract
A new wave of Connected Automated Vehicles (CAV) will put unprecedented strain on current infrastructure. The advancement of infrastructure is necessary to accommodate the best performance of new applications such as Vehicle-to-Infrastructure communication (V2I). One of these challenges is merging into highways where platoons of vehicles drive with short headways which the gaps may become scarce. To enhance the merging for CAVs in highways, this research introduced a simulation-based solution. The proposed method contributes to the design of an automated merging system using V2I communication between vehicles and roadside units (RSU) to make travel time reliable and speed uniform. The designed algorithm assists merging of platoons into a highway adaptively to create gaps that otherwise would not be available. The simulations were designed for a base case without any controller and for several alternatives including slow down on the mainline or on-ramp, and lane change on the mainline. The programming approach is an integration of Python scripted in Vissim. This is a recent method to simulate a connected vehicle environment. For a desired speed of 60 km/h, the average on-ramp speed was 55% higher for the scenarios with the controller compared to the scenario with no controller. The results indicate that for long platoons on the mainline, the automated merging controller improved on-ramp travel time by 74%. Moreover, the controller for average platoon sizes of 6 successfully helped vehicles to merge without disturbance on the mainline by synchronizing gaps availability and on-ramp vehicles arrival to the auxiliary lane.
The Development of a State-Based Fundamental Diagram for Signalized Intersection Using Connected Vehicle Trajectory
Xiaoyu Guo, Texas A&M Transportation InstituteShow Abstract
Xiao Xiao, Texas A&M University
Yunlong Zhang, Texas A&M University
Chaolun Ma, Texas A&M University
Shanglu He, Nanjing University
The fundamental diagram is to describe the relationships between traffic flow and density macroscopically. It is widely used in the traffic analysis for freeways and urban streets. In past studies, the aggregated empirical measurements from detectors are commonly used to fit the diagram. In this study, a set of connected vehicle trajectory input is used to construct the fundamental diagram (i.e. flow-density relation) based on the states the traffic experiences at a signalized intersection. A connected vehicle trajectory method is introduced following the process of data filtering and clustering, critical point extraction, state identification, shock wave formation, and computation of a state-based fundamental diagram. The method is then validated with VISSIM trajectory data and is demonstrated to have a consistent efficiency over different penetration ratios. Lastly, by comparing with detector-based method, the connected vehicle trajectory method generates a better-shaped state-based fundamental diagram at 10% market penetration. Although connected vehicle trajectories are microscopic records of individual driving states, they can interpret the interactions between vehicles (i.e., shock waves). Through this study, the connected vehicle trajectory demonstrates its feasibility and effectiveness to detect the shifts between traffic states, reveal transition traffic interactions and provide the potential to model macroscopic patterns of traffic under low penetration ratios. With the development of emerging connected vehicle technologies, the potential applications of this CV trajectory method will benefit traffic flow modeling and the development of traffic management strategies.
The LWR Model with a Stochastic Speed–Density Relation
Alicia Alcoba, University of California, IrvineShow Abstract
Irene Martinez, University of California, Irvine
Jorge Laval, Georgia Institute of Technology (Georgia Tech)
Wenlong Jin, University of California, Irvine
Traffic oscillations are generated even if the road does not experiment any geometrical change. The formation and propagation of oscillations in a homogeneous road is due to the evident randomness observed in empirical data, which may be generated by drivers’ behavior, acceleration and deceleration processes, vehicles, roads and environmental conditions. This paper proposes a stochastic LWR model that succeed in represent the instabilities observed in the experimental field. The model introduces a stochastic term in the deterministic model definition and solves the problem by converting it to an equivalent car-following model in Lagrangian coordinates. This simple model is validated using Monte Carlo simulation method, which allows to prove their convergence with the deterministic and theoretical approach. Moreover, the convergence analysis stands out the trend of the standard deviation of the speed and allows us to define a formulation for any following vehicle. The calibration of the model is made by analyzing empirical data and proving the match between the stochastic model and experimental observations. Also, the main stochastic parameter is calibrated by formulating an accurate dependency on the free-flow speed of the model. Further studies will introduce bounded acceleration to prove the reliability of the model.
Optimal Fleet Management for Real-Time Ridesharing Service Considering Network Congestion
Negin Alisoltani, Université Paris-EstShow Abstract
Ludovic Leclercq, Universite de Lyon
Mahdi Zargayouna, Université Paris-Est
Jean Krug, Universite de Lyon
When assessing the dynamic ride-sharing problem, two important points should be considered. First, how the ride-sharing system serves the network demand and second, how the ride-sharing system is impacted by the network and in particular by congestion. Most of the existing approaches focus on the first point, i.e. designing the demand matching while using basic assumptions for the second point, mainly constant travel times. Furthermore, most assume that predicted travel times used for the demand-matching are observed when executing the vehicle schedule, which is usually not the case in practice. In this paper two models are defined to deal with dynamic traffic conditions: current mean speed in the network is used over the next 10 minutes to predict travel times when calculating the optimal schedule for the ride-sharing fleet. This fleet is assumed composed of autonomous cars to avoid considering constraints about the drivers. Then, cars travels are simulated and the traffic situation is updated every 10 seconds using a trip-based MFD model as the plant model to represent the traffic dynamics. Some important details are discussed: improvements in the objective functions and also traffic conditions with different values for the number of sharing, the market-rate, and pickup/drop off time window. We find out that the proposed system is really efficient in terms of reducing congestion, especially in peak hours if sufficient sharing happens. Also it can reduce the providers cost while it has small increase in passengers waiting time and travel time.
Modeling and Integrated Control of Macroscopic Heterogeneous Traffic Flow in a Large-Scale, Urban Network Using a Colored Petri Net
Hui Fu, Guangdong University of TechonologyShow Abstract
Kaiyu Chen, Guangdong University of Technology
Saifei Chen, Guangdong University of Technology
Anastasios Kouvelas, ETH Zurich
Nikolas Geroliminis, Ecole Polytechnique Federale de Lausanne (EPFL)
The evaluation of perimeter control using Macroscopic Fundamental Diagram (MFD) concept is strongly related with the quality of the corresponding dynamic models. For capturing the real characteristics of traffic dynamics, an enhanced accumulation-based traffic model is proposed in which transfer flow and travel delay are considered simultaneously using coloured Petri net. Taking the advantage of graphical structure, the gated links and junctions on the border of the protected network are modeled as so-called buffers. Moreover, a perimeter control framework integrated with route guidance is proposed for enhancing the ability of perimeter control on alleviating total travel delay out of the protected network. Firstly, perimeter control inputs are optimized by the method of model predictive control at region level. Secondly, a set of internal flow controllers are adopted to homogenize traffic density among subregions and route guidance strategies are used by monitoring the number of queuing vehicles in buffers. Finally, the proposed traffic model is served as a plant for evaluating network performances objectively. The numerical results clearly demonstrate the effectiveness of our proposed integrated control framework.
A Macroscopic Traffic Flow Model That Includes Driver Sensitivity to the Number of Free Spaces Ahead
Ismael M. Pour, Sharif University of TechnologyShow Abstract
Habibollah Nassiri, Sharif University of Technology
This paper addresses the first-order extension of the Lighthill-Whitham-Richards (LWR) macroscopic traffic flow model. Although previous studies have focused on the fluid aspect of traffic flow, none have addressed the sensitivity of drivers to the number of free spaces within a certain distance ahead of the subject driver. To incorporate driver behavior, we used the number of free spaces ahead of subject drivers and their sensitivity to the number of free spaces within a certain distance ahead. The resulting model is a convection-diffusion model. By computing Einstein's diffusion equation and comparing it with the diffusion coefficient in the extended model, a theoretical relation for the driver's sensitivity was derived. A representation of the numerical results of the LWR model, the convection-diffusion model with a typical diffusion coefficient, and the extended model showed that the proposed model has a lower mean relative error value.
Calibrating Car-Following Models on Surface Roads Using Shanghai Naturalistic Driving Data
Xuesong Wang, Tongji UniversityShow Abstract
Linjia He, Tongji University
Meixin Zhu, Tongji University
Chen Chai, Tongji University
Car-following models are the core component of microscopic traffic simulation, intelligent transportation systems and advanced driver assistance systems. Through analyzing driving behavior characteristics, the models are critical to road design and traffic management. Lack of reliable traffic data in China, however, has limited the use of car-following models, particularly for surface roads. Yet infrastructural issues such as inadequate road design, poor maintenance, and insufficient traffic management have exacerbated the problems caused by mixed traffic flow on surface roads, which is composed of a complex mix of motor vehicles, non-motorized vehicles, and pedestrians. To address this need, five typical car-following models were calibrated and validated with 4,400 surface road car-following events extracted from the 161,055 km of data collected in the Shanghai Naturalistic Driving Study (SH-NDS). The models were evaluated based on their parameter estimates and the Root Mean Square Percentage Errors (RMSPE). Results showed that 1) the Intelligent Driver Model, with a calibration error of 24% and a validation error of 28%, performed best in modeling drivers’ car-following behavior on surface roads; 2) in comparison to car-following behavior on expressways, drivers on surface roads tend to assume a relatively low car-following speed of about 28km/h, and maintain a slightly higher time headway and maximum acceleration. Due to the IDM’s demonstrated high performance on expressways and surface roads, it is reasonable to assume the model may be used to analyze other roadway types.
Evaluating Safety with Automated Vehicles at Signalized Intersections: Application of Adaptive Cruise Control in Mixed Traffic
Ramin Arvin, University of Tennessee, KnoxvilleShow Abstract
Asad Khattak, University of Tennessee, Knoxville
Jackeline Rios-Torres, Oak Ridge National Laboratory
Automated Vehicles (AV) can potentially improve the performance of transportation system by reducing human errors and enhancing mobility and safety. In lower levels of automation, humans are still controlling the vehicle and receiving some advisory information regarding their surrounding environment. This paper investigates the safety impact of low and high levels of AVs, and their interaction with conventional human-driven vehicles at intersections. In order to enhance our understanding of the future interactions between conventional vehicles with AVs, we developed a framework to simulate the mixed traffic environment. Different market penetration scenarios are simulated using VENTOS (VEhicular NeTwork Open Simulator) software. A modified Wiedemann car-following model was calibrated to represent the behavior of human drivers of both conventional and low-level AVs, while an Adaptive Cruise Control model was used to represent high-level AVs. Notably, this study investigates purely the automation improvement in the system. We modified the acceleration and deceleration regimes of the Wiedemann model and tuned it by harnessing real-world connected vehicle data. Next, the simulation was calibrated utilizing two measures of driving volatility to ensure it closely represents the volatility measures of the real-world data. To evaluate the safety performance of a representative intersection under different scenarios, the number of longitudinal conflicts and driving volatility are used as surrogate safety indicators. The results reveal that increase in market penetration rate of LAVs and HAVs substantially improves intersection safety performance by reducing the number of conflicts and driving volatility.
Variable Speed Limit Control for Mixed Powered Two-Wheelers and Cars Traffic
Sosina Mengistu Gashaw, EURECOMShow Abstract
Jérôme Härri, EURECOM
Paola Goatin, Inria Sophia Antipolis
This paper proposes a Variable Speed Limit (VSL) control strategy for a mixed flow of cars and Powered Two-Wheelers (PTWs). Due to the difference in physical and maneuvering characteristics of PTWs and cars, their impact on the traffic flow dynamics is different. Therefore, a control measure adapted to each vehicle class is required. Accordingly, we propose a vehicle-class specific VSL control scheme that regulates the speed limit for each vehicle class according to traffic efficiency and safety objectives, namely minimizing the total travel time and the speed difference between the two vehicle classes, respectively. The dynamics of the mixed traffic flow is formulated in Lagrangian coordinates, which allows vehicle-class specific group/platoon based speed limitations. The proposed VSL control scheme is analyzed through simulation experiments. The results show that vehicle specific control is beneficial both in terms of traffic efficiency and safety.
Predictive Speed Harmonization in a Connected Environment: A Machine Learning Approach
Amr Elfar, Northwestern UniversityShow Abstract
Alireza Talebpour, Texas A&M University
Hani Mahmassani, Northwestern University
Speed harmonization is an active traffic management strategy used for delaying the onset of traffic flow breakdown and mitigating congestion. It changes the speed limits throughout a roadway segment of interest based on prevailing traffic, weather, and road conditions. Implementations rely on fixed roadway sensors to collect traffic information and variable speed signs at fixed locations to display updated speeds. Moreover, most use a reactive rule-based decision tree to activate the control strategy. This setup faces three main challenges: 1) fixed infrastructure sensors do not provide a complete picture of traffic conditions and therefore can impact the accuracy of locating congestion, 2) communicating speed limit changes to drivers at fixed locations can result in an ineffective response depending on where the sign is located with respect to congestion, 3) reactive speed harmonization strategies are generally less effective than predictive ones . To overcome these limitations, this paper presents a predictive speed harmonization system that utilizes machine learning algorithms and V2I communications. The system collects detailed trajectories from connected vehicles to estimate current traffic properties, predict future traffic state, and broadcast new speed limits to connected vehicles accordingly. Simulation of multiple operational scenarios shows that the system can improve traffic flow stability, mitigate congestion, an reduce travel time. The system can also prevent traffic flow breakdown entirely in low traffic congestion conditions. Results also indicate that an optimal speed harmonization strategy requires control at the individual vehicle level where only a small percentage of vehicles need to update their speed.
A Unifiable Multi-Commodity Kinematic Wave Model for Traffic Systems with Tradable Right-of-Way
Pratiik Malik, University of California, IrvineShow Abstract
Wenlong Jin, University of California, Irvine
Roger Lloret-Batlle, University of California, Irvine
R. Jayakrishnan, University of California, Irvine
In this study, we are concerned with the traffic flow system on a multi-lane road, where vehicles with heterogeneous values-of-time (VOTs) can trade their rights-of-way (ROWs) so as to minimize individuals’ travel costs. The resulting traffic flow violates the First-In-First-Out (FIFO) principle, since vehicles with higher VOTs would travel faster by paying those with lower VOTs. A novel multi-commodity kinematic wave model is developed for such a system based on the following five assumptions: (A1) Travel speeds of vehicles not participating in the trade scheme and average speed of all vehicles in the system are not impacted; (A2) The total budget is balanced; (A3) The system reaches user equilibrium, and no driver can reduce his/her cost by unilaterally change his/her choice in speed and payment; (A4) The total cost is minimized, and the system reaches the system optimal state; (A5) The benefits of the scheme are shared among all users. From these assumptions we first derive a unifiable fundamental diagram, in which the relative speed ratios of different commodities are constant and proportional to the square root of the VOTs. We show that the scheme is always pareto-improving when the system optimality is achieved. We analytically solve the Riemann problem for a traffic system with three commodities, in which two groups of users with different VOTs participate in the scheme, and a third group does not participate. We demonstrate that different commodities would react differently to shock and rarefaction waves. Finally we conclude the study with future extensions.
Characterizing Traffic Flows with Mixed Autonomous and Human-Driven Vehicles
Reza Mohajerpoor, University of SydneyShow Abstract
Mohsen Ramezani, University of Sydney
Presence of autonomous vehicles (AVs) affects traffic flow characteristics of a mixed traffic stream comprising human-driven vehicles. In particular, AVs can increase the saturation flow of arterial roads and freeways. To model this impact, we propose analytical models to derive the expected value and variance of headway of a traffic stream with mixed AVs and conventional human-driven vehicles (NVs), given the expected penetration rate of AVs. In addition to modeling the average headway under the general arrangement of AVs and NVs, the lowest and highest achievable headways and their variability are also modeled. We demonstrate that the average headway as a probability function exhibit linear characteristics with respect to the penetration rate of AVs, which helps attain simplified formulas for the expected headway estimation. Microsimulation studies demonstrate the validity of the developed models for the average headway and its variability. The modeling results highlight that: (i) the expected average headway of the mixed traffic under the general arrangement of vehicles is closer to the worst arrangement than the best one for various expected penetration rates, and (ii) the stream's average headway is not sensitive to moderate variabilities of headways between different vehicle types and AV's penetration rate around their nominal expected values.
Statistical and Machine Learning Based Anomaly Detection Using Traffic Detector Data
Dong Pan, George Washington UniversityShow Abstract
Wei Zhang, Federal Highway Administration (FHWA)
Samer Hamdar, George Washington University
In connected travel environments, there is a need of processing information collected from roadway infrastructures and distributing real-time traveler information to drivers for proactive transportation management. Given that traffic detectors are widely used by different transportation agencies and such devices are accessible as a source of descriptive traffic dynamics information, the objective of this study is to utilize these sensors for traffic anomaly detection. Towards realizing such objective, various statistical and machine learning methods are applied and evaluated, including t-statistics, one-class support vector machine (OCSVM) and multilayer perceptron (MLP) neural network. The results demonstrate the feasibility of these methods for traffic flow pattern recognition and thus anomaly detection, and collision identification. Basic t-statistics followed by additional filtering techniques performs better than OCSVM and MLP when the term of traffic anomaly specifically refers to collisions (anomalies are relative to patterns seen at given locations and may be related to weather events, work zones, etc.). For future research, more dedicated machine learning algorithm, for example, recursive neural network (CNN) with long short-term memory (LSTM), is needed for further leveraging traffic detector data to proactively predict and control vehicular flow during disruptive events.