Evaluation of Traffic Delay Reduction from Automatic Workzone

Evaluation of Traffic Delay Reduction from Automatic Workzone Information Systems Using Micro-simulation Lianyu Chu CCIT, UC Berkeley Hee Kyung Kim UC Irvine Henry X. Liu Utah State University Will Recker UC Irvine 1

OUTLINE • Introduction • Methodology • Model calibration • Evaluation • Conclusion 2

Introduction • Work zone – Noticeable source of accidents and congestion • AWIS: – Automated Workzone Information Systems – Components: • Sensors • Portable CMS • Central controller – Benefits: • Provide traffic information • Potentially – Improve safety – Enhance traffic system efficiency 3

Introduction (cont. ) • AWIS systems in market – ADAPTIR by Scientex Corporation – CHIPS by ASTI – Smart Zone by ADDCO Traffic Group – TIPS by PDP Associates • Evaluation studies – Most • System functionalities • Reliability – Few • Effectiveness on safety, diversion 4

Objective & Approach • Effectiveness of AWIS in traffic delay reduction • Methods – Field operational test • Uncontrolled factors, e. g. incidents, variations of demands – Traffic analysis tools • Freq, Quick. Zone, QUEWZ – Difficulty in representing complicated traffic congestion analytically – Proposed method: • Microscopic simulation: Paramics – Model vehicles in fine details 5

Study site and CHIPS system • • Site Location – City of Santa Clarita, 20 miles north of LA – On I-5: 4 -lane freeway with the closure of one lane on the median side – Construction zone: 1. 5 miles long – Parallel route: Old Road – Congestion: occurred in Holidays and Sundays CHIPS configuration – 3 traffic sensors – 5 message signs 6

System Setup Queue Detector Scenario CMS Combo Message RTMS-1 RTMS-2 RTMS-3 PCMS-1 PCMS-2 PCMS-3 PCMS-4 SBS 01 F F F CMB 01 SBS 02 T F F CMB 02 CMB 03 CMB 05 SBS 03 T T F CMB 06 CMB 07 CMB 03 CMB 10 SBS 04 T T T CMB 06 CMB 07 CMB 08 CMB 09 PCMS-5 CMB 11 T = Queue being detected, F = No queue being detected • • • CMB 06 : SOUTH 5/TRAFFIC/JAMMED, AUTOS/USE NEXT/EXIT CMB 07 : JAMMED/TO MAGIC/MOUNTAIN, EXPECT/10 MIN/DELAY CMB 08 : JAMMED/TO MAGIC/MOUNTAIN, EXPECT/15 MIN/DELAY CMB 09 : JAMMED TO MAGIC MTN, AVOID DELAY USE NEXT EXIT CMB 11: SOUTH 5 ALTERNAT ROUTE, AUTOS USE NEXT 2 EXITS 7

Evaluation methodology - Before-after study: 8

Building network • Based on – aerial photos – geometry maps • inputs: – roadway network, – traffic detection, – traffic control, – vehicle data, – driving behavior – route choice – traffic analysis zones 9

Calibration • • • Calibration: – Adjust model parameters to obtain a reasonable correspondence between the model and observed data – Time-consuming, tedious – Models need to be calibrated for the specific network and the intended applications Methods – Trail-and-error method – Gradient- based and GA Proposed 2 -step method: – Calibrate driving behavior models – Simultaneous estimation of OD matrix and route choice 10

Data collection • Before: May 18 th, 2003 • After: Sep 1 st, 2003 (Labor Day) • Link flows: – 5 on-ramps and off-ramps – RTMS-1 and RTMS-3 – Loop detector station at Hasley Canyon Rd • Probe data – Two routes: • Mainline and the Old road – GPS-equipped vehicles 11

Calibrate driving behavior models • • • Calibrate capacities at major bottlenecks Three parameters: – Mean headway – Drivers’ reaction time – Headway factor for mainline links Trial-and-error method – Choose several parameter combinations – Check their performances - Mean headway = 0. 9 - Drivers’ reaction time = 0. 8 12

Simultaneous estimation of OD and routing parameters • Connected and affected each other • Formulated as s. t. • Solution algorithm: – two-stage heuristic search method – Assisted by Paramics OD estimator • using Paramics modeling engine • inputs 13

two-stage heuristic search method • Stage 1: Generate n sets of OD tables and routing parameters with Paramics OD estimator – (1) Choose n routing parameters θ 1 to θn from multiple dimension parameter space. – (2) Let i = 1, set θ= θi. – (3) Estimate OD table Гi based on traffic counts at measurement locations. The objective function is: – (4) i = i+1. If i < n, go to step 3; otherwise go to the next stage. 14

two-stage heuristic search method • Stage 2: Evaluate these OD tables and routing parameters with Paramics Modeler. – (1) Let i = 1 – (2) Run Paramics simulation with OD table Гi and θi. – (3) Evaluate the estimated OD and routing parameters: – (4) i = i+1. If i < n, go to step 2, otherwise go to Step 5. – (5) Compare all MAPE(i), the lowest one corresponds to i = μ. – Therefore, Гμ and θμ are the best calibrated OD table and routing parameter. 15

Model calibration for step 2 • Route choice model in simulation models – Dynamic feedback assignment – Parameters: feedback cycle, compliance rate • Routing parameter θ: • 1 parameter: compliance rate • Inputs to OD estimator: – Reference OD table from planning model – 6 cordon flows – 5 measurement locations • Simulation period: 3 -5 pm – Warm-up: 3 -4 pm 16

Calibration results • Stage 1: – routing parameter θi : an estimated OD table Гi • Stage 2: 17

Calibrated “after” model Location/Route Traffic count (veh/hour) main_Hasley main_Rye Canyon Observed Simulated APE (%) 3890 4568 3867 4526 0. 59 0. 92 764 480 713 813 477 674 6. 41 0. 63 5. 47 Mainline 13 13. 1 0. 48 Old Road 11 11. 1 0. 92 off_Hasley on_SR-126 on_Magic Mountain Travel time (min) MAPE 2. 20 18

Evaluation • Run two simulation models: – Demand: after OD table – Before: • calibrated before network, 3% compliance rate – After: • Calibrated after network, 18% compliance rate • Number of simulation runs – Median run with respect to VHT 19

Evaluation results 20

Evaluation results (cont. ) 21

Conclusion • a microscopic simulation method to evaluate traffic delay reduction from AWIS – Calibration of two simulation models • two-step calibration – a simultaneous estimation of OD table and routing parameters • A two-stage heuristic solution algorithm • Evaluation: – AWIS can effectively • reduce traffic delay • Improve overall performance of the traffic system 22