PERFORMANCE MEASUREMENT AND MONITORING IN TSM&O - CURRENT PRACTICE AND FUTURE DEVELOPMENTS

Transcription

1 PERFORMANCE MEASUREMENT AND MONITORING IN TSM&O - CURRENT PRACTICE AND FUTURE DEVELOPMENTS Speakers: Tony Kratofil Michigan DOT Aleksandar Stevanovic FAU Michael Pack UMD CATT Lab Sponsored by TRB NCHRP and AASHTO STSMO Performance Measures Working Group NOCoE Webinar, April

2 Webinar Agenda Presentations NCHRP 20 07/Task 366 Accessing Information about Transportation Systems Management and Operations Performance Measurement Aleksandar Stevanovic, FAU LATOM Answering Questions Both Known & Unknown: TSM&O Performance Measurement & Monitoring with Big Data Michael Pack, UMD CATT Lab Two methods to estimate traffic performance in urban networks Aleksandar Stevanovic, FAU LATOM Estimating Signal Performance based on Link Travel Times Estimating Network Congestion based on Google Traffic Maps 2

3 NCHRP 20 07/Task 366 Research Objectives Establish a framework for organizing information about research and practices for TSMO performance measurement and monitoring to assess the impacts of TSMO strategies Less well developed Most difficult to measure Facilitate access to TSMO performance measurement and monitoring information by developing a guide to the most relevant recent literature Identify and describe the problems, opportunities, and consequences for practitioner adoption of specific measures and setting targets for TSMO performance management, with particular attention to federal rulemaking undermap 21 legislation 3

4 Research Approach Develop a methodology to categorize existing performance measurement studies by creating a comprehensive list of potential categories Start from existing categories, i.e., Elements of Success Review and edit existing categories as necessary, and suggest new categories and their subcategories for specificity In collaboration with NCHRP, SHRP2, and TRB, identify the best way to integrate research outputs into either : Existing TSMO web platform ( One or more other existing websites serving the intended audience A brand new, custom designed website Discuss an analysis of the various problems, opportunities, and consequences of the adoption of each measure Reach out to leading performance measures implementers to evaluate the impacts of the methodologies and metrics 4

5 Information Organization Framework and Literature Search Develop Information Organization Framework Develop preliminary categories to classify/filter various performance measurement related studies Propose how to integrate such categories into existing tsmoinfo web portal or into a new web site Conduct Literature Search Identify ongoing and past research efforts relevant to the project Create a list of studies and the categories (elements of success) 5

6 Former tsmoinfo.org Structure 6

7 Current transportationops.org Structure 7

8 Considered TSM&O Strategies in this Study TSM&O Strategies 1 Access Management 2 Active Parking Management 3 Active Traffic Management 4 Adaptive Traffic Signal Technology 5 Bicycle and Pedestrian Management 6 Corridor and Arterial Traffic Management 7 Freeway Management 8 High Occupancy Vehicle (HOV) Lanes 9 Pricing/Toll Roads 10 Ramp Metering 11 Geometric Design 12 Traffic Signal Program Management 13 Signal Timing 14 Transit Operation 15 Transit Signal Priority 16 Travel Demand Management 17 Freight 18 Road Weather Management 8

9 TSM&O Performance Measures Categories Performance Measures Operations Safety Environment Economics 9

10 Descriptions of Topic Indexing Methods Method Alternative Name Working Definition Very specific general categories, like Planning or Text categorization Text classification Operations, are assigned from usually a small vocabulary in the context of performance measures. Term assignment Subject indexing Main topics are expressed using terms from a large vocabulary, e.g. a thesaurus. The list of categories created in this Task 2, can serve as our thesaurus Key phrase extraction Keyword extraction, Key term extraction Main topics are expressed using the most prominent words and phrases in a document Terminology extraction Back of the book (BOB) indexing All domain relevant words and phrases are extracted from a document Full text indexing Full indexing, Free text indexing All words and phrases, sometimes excluding stop words, are extracted from a document Key phrase indexing Full indexing, Free text indexing All words and phrases, sometimes excluding stop words, are extracted from a document Key phrase indexing Key phrase assignment A general term, which refers to both term assignment and key phrase extraction Tagging Collaborative tagging, Social tagging, Auto tagging, Automatic tagging The user defines as many topics as desired. Any word or phrase can serve as a tag. Applies mainly to collaborative websites 10

11 Framework for Retrieval of Information 11

12 Finalized Organizational Framework 12

13 Example of Performance Measures Categorization 13

14 Example of Literature Overview 14

15 Problems and Opportunities for Common TSMO Specific Performance Measures A comprehensive list of performance measures for various subareas of TSMO A set of matrices which will be used to categorize these performance measures according to various aspects of deployment 15

16 Example of TSM&O Performance Measures TSMO Strategy Sub category Performance Measure Travel Time Operation Speed Delay Queue Length Crash Rate (Crash per million VMT) Access Management Safety Economic Number of Crashes Number of Injury and Fatalities Number of Property Damages (Number of property damages only (PDO)) Time to Collision (Time takes for a vehicle to collide into another if they continue at the same speed without trying to avoid each other) Business Turnover Commercial Land Values Property Value 16

17 Use of Various Performance Measures Performance Measures Organization Sub category: Operation Travel Time Reliability (Variation in travel time) NCFRP, FHWA VDOT, ODOT, NYSDOT, CDOT, University Transportation Center for Alabama, Travel Time Index NCDOT, ODOT, FDOT, FHWA, Delaware Valley Regional Planning Commission, Buffer Time Index FHWA, Delaware Valley Regional Planning Commission, Planning Time Index Delaware Valley Regional Planning Commission, Sub category: Safety Crash Rate NCHRP, FHWA, ODOT, VDOT, NCDOT, TxDOT, FDOT, DDOT Crash Severity ODOT, University Transportation Center for Alabama, Fatality Rate NCFRP, TxDOT, ODOT, FDOT, DDOT, NCHRP Rate of Injuries VDOT, Sub category: Economic Business Turnover VDOT Commercial Land Values VDOT Property Value NCHRP, FDOT, ODOT, Deployment Costs DOT Fuel Consumption Nevada DOT, MnDOT, TxDOT, ITS CT, FHWA Sub category: Environment Emission (hydrocarbons, carbon monoxide, nitrogen oxides and volatile organic compounds) TxDOT, FDOT, NYSDOT, WSDOT, ODOT, University Transportation Center for Alabama, NCHRP, NCFRP, FHWA, Texas Transportation Institute Carbon dioxide Emission TxDOT, FDOT, NYSDOT, WSDOT, ODOT, NCHRP, NCFRP, FHWA Emission of Greenhouse Gas WDOT, FDOT, ODOT 17

18 Example of Performance Measure Matrices Performance Measure Definition Units Spatial Scope Time Scale Sub category: Operational Specific points on a Peak hour, section or a Average time consumed by vehicles travelling a a.m./p.m. peak Travel Time Minutes representative trip; fixed distance period, midday, separate for GP and daily HOV lanes Travel Time Reliability Travel Time Index Speed Delay Measure of dispersion or spread of travel time distribution Ratio of actual travel rate to ideal travel rate Average speed obtained by the vehicles in a fixed distance Excess travel time used on a trip, facility, or freeway segment beyond what would occur under ideal conditions Minutes None; Minimum value = 1.00 Mile per hour Vehicle hours Specific section or a representative trip only Section and area wide as a minimum; separately for GP and HOV lanes Specific points on a section or a representative trip only; separate for GP and HOV lanes Section and area wide as a minimum; separate for GP and HOV lanes Peak hour, a.m./p.m. peak period As needed Peak hour, a.m./p.m. peak period, midday, daily Peak hour, a.m./p.m. peak period, midday, daily 18

19 Data Collection, Advantages and Disadvantages Performance Measures Travel Time Travel Time Reliability (Variation in travel time) Travel Time Index Buffer Time Index Planning Time Index Speed Delay Intersection Delay Congestion Hours Techniques to Collect Data Probe vehicle techniques Instrumentation Level Manual GPS Electronic DMI Advantages Low initial cost No special equipment needed Low required skill level Moderate initial cost Reduction in human error Data easily integrated into GIS Detailed speed/delay data available No vehicle calibration is necessary as with the DMI method Moderate initial cost Proven technology Reduction in human error Very detailed speed/delay data available Commercially available software provides a variety of collection and analysis features Disadvantages High operating cost (high labor requirements) Greater potential for human error Limited travel time/delay information available Limited sample of motorists Reception problems in urban canyons, trees Limited sample of motorists Due to rapidly changing area, difficult to stay updated on what equipment to purchase Not readily adaptable to a geographic information system Limited sample of motorists 19

20 Future Steps Review of Project related MAP 21Efforts MAP 21 has six primary goals which include enhancing safety, improving infrastructure condition, reduction of congestion, increasing system reliability, development of freight movement and economic vitality, and improving environmental sustainability. Identify Impact of MAP 21 Legislation on TSMO specific Performance Measures Review of FHWA MAP 21 congestion related rulemaking efforts, AASHTO activities, and ongoing research and development work 20

21 MAP 21 Goals and Performance Measures Goals Performance Measures Definition Units Serious injuries per VMT Total crashes divided by the total vehicle miles traveled for which a police accident report form is generated, where at least one injury occurred. Persons per mile Fatalities per VMT Total fatal crashes divided by the total vehicle miles travelled, for which a police accident report form is generated, where at least one fatality occurred. Persons per mile Safety Number of serious injuries Total crashes for which a police accident report form is generated, where at least one injury occurred. Persons Number of fatalities Total fatal crashes for which a police accident report form is generated, where at least one fatality occurred. Persons Number of transit related fatalities IRI (International roughness index) Total fatal crashes related to transit system for which a police Persons accident report form is generated, where at least one fatality occurred. Ride Quality Parameter (IRI) IRI is the International Roughness Index and measures pavement smoothness. m/km or mm/m (from 0 to 170) Infrastructure Conditions Pavement structural health index Percentage of pavement which meet minimum criteria for pavement faulting, rutting and cracking. Percentage 21

22 Answering Questions Both Known & Unknown: TSM&O Performance Measurement & Monitoring with Big Data Michael Pack, UMD CATT Lab 22

23 Estimating Signal Performance based on Link Travel Times 23

24 Purpose of Research Many agencies deploy ITS equipment to measure arterial travel times Very few studies (if any) investigated if travel time information can be used to estimate performance of traffic signals We present a method to estimate performance of traffic signals (their major coordinated movements) based on point to point travel time measurements The core of this method is based on well known volume delay functions which have been used in transportation planning for decades Use of these relationships has been reversed to estimate some fundamental signal performance measures of the downstream signal (e.g. V/C ratio, Level of Service (LOS), number of cycles to pass through the signal) based on travel time between pairs of signalized intersections 24

25 Overall Framework Input Data Upstream travel time Volume delay function (VDF) to establish this relationship Signal Performance Measures Volume to capacity ratio Level of service Number of cycles 25

26 Defining Travel Time V/C Relationship 1) Retrieving travel time from Acyclica (or BlueTOAD) 2) Retrieving signal timing parameters from ATMS.now Aggregating travel time by every cycle 3) Traffic volume count using CCTV Count # of vehicles through/queued vehicles 4) Estimating capacity and saturation flow rate Estimating hourly volume using the observed # of vehicles and cycle length Saturation flow rate = capacity * (effective green time/cycle length) 5) Estimating V/C Retrieve occupancy rates from Sensys and find free flow conditions Defining free flow travel time 6) VDF Curve plotting and calibration MATLAB Curve fitting toolbox finds the best combinations of calibration parameters 7) VDF validation Validate the calibrated VDF with a new set of data collected a different link 26

27 Study Area & Data Collection 27

28 Data Collection: Volume Volume = V 1 (Passing Vehicles) + V 2 (Queued Vehicles) (Passing Vehicles) Counting passing vehicles during at the stop line (Queued Vehicles) Counting queue at the beginning of 28

29 Data Collection: Acyclica Travel Time User Interface & Output Data Format 29

30 VDF Formulas Used Category VDF Formula Newly developed Conventional (Most common VDFs) Conventional (VISUM) Bureau of Public Roads (BPR) Conical Akcelik. BPR2 BPR3 Conical_Marginal Logistic Quadratic t = Exponential Inrets.. Lohse.. 30

31 Example: Data Points Collected 400 NW 10 th Int. to Airport Rd. 313 Data Points 350 Travel Time (Seconds) V/C 11 Feb 16 Feb 17 Feb 18 Feb 31 Mar 1 Apr 31

32 Example: VDF Parameter Optimization Occupancy Rate (%) Optimizing Parameters Using MATLAB Curve Fitting Toolbox 0 Time April 1st April 2nd April 3rd 45% Frequency (%) 40% 35% 30% 25% 20% 15% 10% 5% 0% Optimization Algorithm Least square Method Bins (Travel time in seconds) 32

33 Example: VDF Parameter Optimization VDF Calibration and Plotting Results New VDF & BPR (Calibrated) Other VDFs Travel Time (Seconds) V/C Observed Newly derived function Calibrated BPR Travel Time (Seconds) V/C Observed BPR BPR2 BPR3 Conical Conical_M Akcelik Logistic Quadratic Exponential Inrets Lohse 33

34 Example: Calibrated VDF Parameters VDF Functions Calibrated Parameters c d f RMSE R squared BPR BPR BPR Conical Conical_Marginal Akcelik Logistic Quadratic Exponential Inrets Lohse New VDF

35 VDFs on Glades Rd Free flow T.T.: 25 s BPR (R 2 = 0.55).. New VDF (R 2 = 0.57) Free flow T.T.: 25 s BPR (R 2 = 0.59).. New VDF (R 2 = 0.58) Free flow T.T.: 25 s BPR (R 2 = 0.70).. New VDF (R 2 = 0.71) Free flow T.T.: 12 s BPR (R 2 = 0.79).. New VDF (R 2 = 0.79) Free flow T.T.: 50 s BPR (R 2 = 0.78).. New VDF (R 2 = 0.78) NW 13 th St. 2 NW 10 th Ave. 3 Airport Rd. 4 I95 NB Off ramp 5 I95 SB Off ramp 6 Renaissance Way 7 Butts Rd. 8 Town Center Mall Entrance 9 St. Andrews Blvd 35

36 V/C Estimation Simulation TT8 (291.0, 1.42) (358.8, 1.59) Travel Time (sec) Estimated V/C TT TT6 (232.4, 1.29) (190.0, 1.16) TT5 TT4 (157.8, 1.04) TT3 (110.0, 0.82) TT2 TT1 (39.7, 0.29) (75.5, 0.56)

37 V/C Estimation Simulation % 100% % 30% 42% 10% 2% 100% % 33% 45% 7% 100% % 21% 54% 24% 60% 40% % 90% 21% 63% 16% % 93% 7% 76% 17% % 69% 31% % 50% 25% 25% LOS is F in 90% and E in 10%. The vehicles are expected to pass the intersection within 1 cycle (21%), 2 cycles (63%), and 3 cycles (16%). 37

38 Visual Validations 38

39 Practical Uses V/C Ratio Oversaturation Near Saturation Normal Travel Time (Seconds) Oversaturation 1.00 Near Saturation V/C Ratio Normal Travel Time (Seconds) 39

40 Current Interface & Demonstration Functionalities & Demonstration Only using the upstream travel time Estimating three signal performance measures V/C Level of Service Number of Cycles Showing the summarized information (Default Setting) Showing the detailed information by clicking the Info. Box. 40

41 Estimating Network Congestion based on Google Traffic Maps 41

42 Assessing congestion based on Google traffic maps Application based on Google traffic maps Initiated from a question can we use information from Google maps to assess network wide congestion? Benefits for TMCs that do not have enough ITS infrastructure A simple idea: program counts # of pixels of certain color and compares with total # of pixels in a specific link Links are defined manually (only once) 42

43 Program Architecture View the map Trigger the (pixel) analysis process Indicate the results on the interface Create a map using latitude/longitude obtained Insert time interval for monitoring Generate output files Load a map and open it in a browser Selected links in the link list? & Time interval inserted? No User decides when to stop Wait until first screenshot is taken Yes Insert time interval for monitoring Create/save links, intersections, and corridors on the screenshot Start the analysis 43

44 Color Scheme & Outputs Google Color Scheme Legend Colors Green Yellow Red Black Traffic Condition Normal Operation Moderate Congestion High Congestion Severe Congestion Output File Format Date Time Link # of Total Pixels captures # of Pixels for Green # of Pixels for Yellow # of Pixels for Red # of Pixels for Black %of Green % of Yellow % of Red % of Black 44

45 Current Interface & Demonstration Selecting part of the map and saving it as a new map. Creating multiple links/ intersections/corridors by drawing on the map. Adjusting the data collection interval. Setting the congestion warning threshold. Detecting the congestion level by capturing the number of pixels (Black/Red/Yellow/Green) Creating output files 45

46 Defining Congestion Threshold Threshold reached 55% of this link is very congested 46

47 Textual Outputs of the Congestion 47

48 Thank You! Questions & Comments? Aleks Stevanovic, Michael L. Pack, Visit 48