V2X Simulation and Modeling Martin Treiber TU Dresden Oct 2010, Dagstuhl
General: About what I will talk – and about what not Models Level of Detail Applications Ray tracing/wave propagation Networking models (NSx) Path models Car-following models Fluid-dynamic models Physical details Packet transmission details Transmission strategy Microscopic traffic flow Macroscopic traffic flow Safety Efficiency Entertainment
Longitudinal communication path Overview General Longitudinal communication path - Assumptions and analytical models - Integration in a microscopic traffic simulator - Results: Which minimum equipment rate is necessary? Transversal (store-and forward) communication path - Analytical models - Simulations and minimum equipment rates - Discussion: Which strategy for which application? Applications for enhancing efficiency - Congestion warning system (V2V / I2V) - Traffic flow assistant (V2V / I2V) - Traffic light assistant (I2V)
Transmission Strategy: V2Vlong, V2Vtrans, V2I, and I2V
Longitudinal Communication Path: Assumptions Direction of communication Driving direction Finite broadcast range r: Either fixed or statistically distributed Vehicle density ρ: Either constant or inhomogeneous (stop-and-go traffic) Constant partial density of equipped vehicles: λ=αρ Poisson-process for the positions of “nodes”, i.e, distances between nodes are exponentially distributed Instantaneous, error-free message transmission, if within range
Longitudinal Communication Path: Analytical Model Poissonian assumption: f(y)= Probability of availability at distance x from sender: prob density avail. at x-y Initial condition: Closed-form analytical solution: Dousse, Thiran, Hasler, Connectivity in Ad Hoc and Hybrid Networks. IEEE INFOCOM Vol. 2, pages 1079-1088 (2002).
Predictions of the Analytical Model Variable range Variable equipment rate actual Potential Length of communication chain chain Result: Longitudinal strategy is not efficient as long as broadcast range R is lower than average distance 1/λ length
Test of Fixed Range Assumption: Deterministic vs Test of Fixed Range Assumption: Deterministic vs. Stochastic Range Model
Robustness Test 1: Deterministic Analytical Model vs Robustness Test 1: Deterministic Analytical Model vs. Simulated Trajectories (Gipps Model) Nearly perfect agreement for free traffic and rates below 30% => small l Simulated connectivity slightly higher for homogeneous congested traffic Simulated connectivity slightly lower for longer distances and stop&go traffic
Robustness Test 2: Analytical Model vs. Real Trajectories (NGSIM Data) First 20 min Jam Spatiotemporal Density Real gap distribution (across all lanes) Free traffic Last 10 min (jam) First 20 min Crossover!
Transverse Communication Path: „Store & Forward“ Communication Connectivity Statistics of Store-and-Forward Intervehicle Communication, IEEE Transactions on Intelligent Transportation Systems 11(1), 172-181 (2010).
Analytical Model for Store & Forward Communication Problem statement: Message has to propagate in upstream direction (at least) distance x Time T1 to find a relay car: Time T2 to reach connectivity to the target region: t*=(x-2r2)/V2 Time T3 to find a car in the target region: Schönhof, Kesting, Treiber and Helbing, Coupled vehicle and information flows: Message transport on a dynamic network. Physica A 363, p. 73-81.
Predictions of the Analytical Deterministic Model Transversal hopping has robust connectivity already for small penetration rate α, i.e., small partial density l=raI A successful transmission is only a matter of time Flip side: The transmission is not instantaneous r1=r2=30 Veh./km, V1=V2=90 km/h, 1 lane
Robustness Test 1: Deterministic vs. Stochastic Model Deterministic: Fixed range R=200 m Stochastic: Exponentially distributed range, E(R)=200 m r1=r2=30 Veh./km, V1=V2=90 km/h, 1 lane
Robustness Test 2: Violation of the Poissonian Assumption (for 1 and 2 Lanes) Solid lines: Analytial deterministic model Symbols: IDM/MOBIL microsimulations
Implementation and simulation of inter-vehicle communication Message: Data structure EquippedVehicle: Mailbox with send() and receive() MessagePool: “Ether” for gathering and distributing Messages Simulate!
Applications for Enhancing Efficiency: Problem Statement Capacity of road network is limited Construction of new roads not feasible due to cost/land constraints How can we avoid / reduce traffic congestion ? Vehicle-based Intelligent Transportation Systems ! 17
Application 1: Congestion warning system (V2Vtrans)
Basic Communication Scheme Scenario: Evolution of congested traffic Trajectories of equipped cars (α=3%) Equipped vehicles generate messages Online traffic state estimation based on floating car data.To appear in Traffic and Granular Flow '09, Springer (2009).
Enhancing the Jam Front Prediction Using V2I, I2I, I2V RSU down RSU up
Simulation Results
ACC-Based Traffic- Flow Assistant Application 2: ACC-Based Traffic- Flow Assistant World 2-3 fahrzeuge Kesting, Treiber, Helbing : Enhanced Intelligent Driver Model to Access the Impact of Driving Strategies on Traffic Capacity. Philosophical Transactions of the Royal Society A (2010). 22 22
(1) The “Ordinary” ACC The Model The System World 2-3 fahrzeuge 23 23
(2) From Conventional to Traffic-Efficient ACC Automated “traffic-efficient” driving strategy Strategy adapts driving style to surrounding traffic situation Do not change model but model parameters Driving strategy matrix in relative terms (preserving individual settings) Kesting, Treiber, Schönhof, Helbing : Adaptive Cruise Control Design for Active Congestion Avoidance. Transportation Research Part C: Emerging Technologies (2008). 24 24
Traffic-Adaptive Driving Strategy in Free Traffic Driving strategy for free traffic and bottleneck situations: Traffic state Driving behavior λT λa λb Free traffic Default 1 Bottleneck Breakdown prevention 0.7 1.5 Bottleneck section: avoid traffic flow breakdown Reduce time gap parameter and increase stability Dynamically “fill-up” capacity gap Increase free capacity Treiber, Kesting, Helbing: Three-phase traffic theory and two-phase models with a fundamental diagram in the light of empirical stylized facts. Transportation Research Part B: Methodological (2010). 25 25
Driving Strategy in Congested Traffic Driving in congested traffic: Safely approaching jam front Increase outflow by quickly accelerating when leaving traffic jam Traffic state Driving behavior λT λa λb Upstream jam front Increased safety 1 0.7 Congested traffic Default / Normal Downstream jam front Increase outflow 0.5 2 Typical sequence of traffic states for a floating car : 26 26
Spatial-Temporal Dynamics of a Freeway Traffic Jam Problem: How to distinguish exiting just a stop & go wave from exiting the whole jammed region? Solution: Knowledge of Traffic flow dynamics and C2X communication! Treiber, Kesting, Wilson: Reconstructing the Traffic State by Fusion of Heterogeneous Data. Computer-Aided Civil and Infrastructure Engineering (2010). 27 27
Realization: “Everything-in-the-Loop” Simulation
The Communications Subsystem
The Simulator Simulate …
Simulation Results: Impact on Collective Travel Time and Fuel Consumption Averaging Problem!
“Traffic-Light Assistant” in City Traffic Application 3: “Traffic-Light Assistant” in City Traffic First Step: Driver information system Traffic light reports about status Wireless communication to cars Cars give recommendations to driver ex: “Travolution” by Audi et al. Next step: Automatic driving Stop line additional “virtual” target In combination with ACC system Aims: Convenience/Safety/Efficiency 32 32
Driving Strategy of the “Traffic-Light Assistant” Basic functionality: Safety / convenience Decision making before light switches Automatic halt at stop line Objectives: Reducing number of stops, fuel consumption, time consumption … “Intelligent” functionality: Efficiency Feature 1: Proactive acceleration (boost) for crossing TL during green light Feature 2: Use motor brake for fuel-optimal approaching / jump-start Coast down Scenario Stops Change Reference 46.1 % Feature 1 40.2 % -13 % Feature 2 38.3 % -17 % Unpublished. © TUD + Volkswagen AG 33 33
Thank you for your Attention www.verkehrsdynamik.de www.traffic-simulation.de www.traffic-states.com 34 34
Supplements
Protection and Prevention is a Success Story German statistics over 40 years: Mileage nearly tripled Fatalities reduced by 75% ! Similar developments in all western countries 36 36
In which way can V2V/V2I make cars smarter? Anti-lock Brake System (ABS) Airbags Electronic Stability Control (ESC) Passenger Protection Pedestrian Protection Adaptive headlights Navigation Vehicle- Infrastructure Integration (VII) Radar-based systems Ultrasonic-based systems Car-to-car communication (C2C) Video-based systems 37 37
Stylized Facts of Spatiotemporal Evolution (A9 South, Germany) Downstream front: Either fixed or moving upstream with velocity c Upstream front: Determined by supply and demand Internal structures: Moving with velocity c as well Amplitude of internal structures grows when moving upstream Frequency grows with severety of bottleneck TSG OCT
Summary: From Driver- to “Traffic-Assistance” Systems Required Information Action Aim ACC Track leading vehicle Control acceleration Driver comfort Traffic-adaptive ACC Bottlenecks and jam fronts ACC parameter adaptation “Flow-efficient” driving strategy Traffic-Light Assistant Traffic light phases Automatic approach of TL “Stop-avoiding” driving strategy Demonstration of automated driving strategies: Implementation in test cars by Volkswagen Positive impact on collective dynamics in simulation Individual benefit from additional driver information 39 39