Green and Smart Transport. Professor Jianping Wu Tsinghua-Cambridge-MIT Future Transport Research Center Tsinghua University, China

Transcription

1 Green and Smart Transport Professor Jianping Wu Tsinghua-Cambridge-MIT Future Transport Research Center Tsinghua University, China 1

2 Contents Researches on Green and Smart Transport 1. Urban emission pollution control with on-line traffic management strategy 2. The driving behavior recognition using Big Data technologies 3. AlphaGo and the future urban traffic control 4. Simulation for Autonomous Driving

3 1. Urban emission pollution control with on-line traffic management strategy

4 Project Team 863 Key Project Research Team: Tsinghua University Peking University Nankai University Nanjing Environmental Monitoring Center

5 Background 2014 ( Beijing) Vehicle Emission Contribution 30-40% PM Million Car Ownership Year

6 Work Packages Work Package1:Internet of Things (IoT) based traffic data monitoring system Work Package2:Flow and speed based emission model Work Package3:Regional multi scale air quality model based on the coupling of meteorological and vehicle emission Work Package4:Traffic management technology based air quality control system Work Package5:System demonstration

7 Operation Flowchart Traffic Flow Data Emission Estimation Air Quality Analysis Traffic Control Needs Control? Web GIS Outputs

8 System Framework 多尺度空气扩散模拟 仿真调控验证 Web 服务器 智能应用客户端 气象站数据中心 数据服务器 尾气排放计算 交通数据融合 RFID 读卡 清华大学移动监测车 环境检测中心

9 The Traffic Flow Data RFID Scanning

10 The Traffic Flow Data GPS Floating Vehicle Data

11 Traffic State and the Emission Vehicle State and Emission 11

12 Traffic State and the Emission Traffic Flow State Emission Model Speed (km/h) Time (s) EURO 5 cycle 1 cycle 2 cycle 3 cycle 4 cycle truck EURO <7,5t 4 truck 7,5-14t truck truck 14-20t EURO <7,5t 3 cycle 1 cycle 2 cycle 3 cycle 4 cycle truck truck 20-28t 7,5-14t cycle 1 cycle 2 cycle 3 cycle 4 cycle truck truck trailer truck 14-20t <20t EURO <7,5t 2 truck truck trailer20-28t truck 7,5-14t cycle 1 cycle 2 cycle 3 cycle 4 cycle truck truck trailer truck trailer truck 14-20t 28-32t <20t EURO <7,5t 1 truck truck trailer>32t truck trailer20-28t truck 7,5-14t cycle 1 cycle 2 cycle 3 cycle 4 cycle semi truck trailer<32t truck truck trailer truck 14-20t 28-32t <20t semi truck trailer truck trailer>32t trailer20-28t Pre <7,5t EURO 1 truck truck 7,5-14t cycle 1 cycle 2 cycle 3 cycle 4 cycle city semi bus truck <8t trailer<32t truck truck trailer truck 14-20t 28-32t <7,5t <20t... semi truck trailer truck trailer>32t truck trailer20-28t truck 7,5-14t... city semi bus truck <8t trailer<32t truck truck trailer 14-20t 28-32t <20t... semi truck trailer truck trailer>32t truck trailer20-28t Emission factors [g/km] for:... city semi bus truck <8t trailer<32t truck trailer 28-32t <20t... semi truck trailer trailer>32t FC, NOx, PM, HC, CO truck trailer20-28t... city semi bus truck <8t trailer<32t 28-32t... semi truck trailer trailer>32t... city semi bus <8t trailer<32t... semi trailer >32t... city bus <8t CO2 [g/h] Driving resistances & transmission losses Emission Map Nm U/min Gear shift model Transient Correction x 3 loadings (empty, 50%, full) x 6 road gradients (-6%, -4%...+6%) Engine load, load, FC, FC, Emissions emissions Cold start tool

13 Traffic 排放与污染分布图 State and the Emission

14 Scale Classification of Atmospheric Diffusion Transport Model Street Scale City Scale Reginal Scale Intercontinental scale 14

15 Pollutant Concentration in Typical Street Canyons 15

16 The relationship between pollutant exposure and mortality Health effects [Schwartz, 2008] 16

17 17 The Demo Site

18 18 Traffic Control Strategy

19 19 Traffic Control Strategy

20 The Simulation Assessment

21 21 Simulation Assessment Results

22 Simulation based Assessment --Air Quality Changes Nanjing, ,9:00-15:00, NOx

23 2. The driving behavior recognition using Big Data technologies

24 Drivers have different driving behavior

25 Purpose Find the Gene Level driving behavior, which is unique to individual driver

26 The Experiment

27 The Experiment EV LV

28 The Experiment

29 The Experiment

30 The Experiment

31 Clustering 1 2

32 Combinational Clustering

33 The Gene Level Driving Behavior

34 The Different Driving Behavior

35 Potential Applications Select Good Drivers Big Data based simulation Provide Personalized Driving Assistance System

36 3. Alpha-Go and the smart signal control system

37 Google DeepMind Challenge Match In March 2016, the World Go Master, Li Shishi, lost to Google s AlphaGo

38 DeepMind- A Trial

39 DeepMind- A Trial

40 DeepMind- Networks Scale-free Network Small World Network Random Network Regular Network

41 DeepMind- Smart Road Network Signal Control System State:Road Network Traffic Flow and other constrains Action: Signal Setting Feedback:Network Total Delay

42 How to verify DeepMind- Smart Road Network Signal Control System The Challengers: High complexity: large-scale road network and traffic state signal value space is extremely large Other constrains Real-time requirements Convergence of algorithm

43 4. Simulation for Autonomous Driving

44 Simulation for Autonomous Driving

45 Thank You!

46 5. On-line simulation and smart transport

47 Conventional Traffic Simulation

48 On Line and Off Line Simulation Conventional (off line)simulation On Line Simulation Platform Computer On PC Embedded into Urban Traffic Management System Road Network Data Input On Site Survey Input from Urban GIS Traffic Flow Data Input On Site Survey Input from On Line Traffic Data Functions 1. Off line Evaluation and Prediction 1. Off Line Evaluation and Prediction + 2. On Line Traffic Management and Control

49 On-Line Simulation for Smart Transport On Line Simulation Assessment Traffic Management Expert System Urban Traffic Data Center Optimized Solution

50 On Line Traffic Management Expert Systems the Optimized Solution On Line Simulatio n Becoming a Decision Making System in Intelligent Transport Systems

51 Typical Application Example 前往城东城中方向车辆, 绕行丰潭路 前往城中方向车辆, 绕行教工路 Beijing Project: Total Travel Time Saving: 1 million hours/day

52 Typical Application Example 4-6 hours on road in average! On Line Simulation Estimation Guidance for Automated Vehicle Dynamic Traffic Guidance

53 The IBM and Geneva Awards

54 City Modeling and Simulation