DCN Z44035C30DES00R02. Advanced Traffic Management System (ATMS) Release 2. Code Specifications (540) - SWARM

Transcription

1 Advanced Traffic Management System (ATMS) Release 2 Code Specifications (540) - SWARM Author: File Name: Revised Date: NET ATMS540_SWARM_FINAL

2 TABLE OF CONTENTS 1.0 INTRODUCTION Purpose of Document Document Organization BASIC INFORMATION Detector Data System Configuration HIGH LEVEL SOFTWARE ARCHITECTURE Data Acquisition (DAQ) Modifications and Additions RMS Modules DAQ MODIFICATIONS AND ADDITIONS DETAILED DESIGN Phase One Failure Management Model Description Inputs and Outputs Pseudo Code Data Normalization and Basic Statistics Description Inputs and Outputs Pseudo Code RMS MODULES Mode Manager Description Dynamic Saturation Density Description Inputs and Outputs Pseudo Code SWARM 1 Algorithm Description SWARM 1 Metering Rate Apportionment and Propagation Pseudo Code LINEAR FORECASTER SWARM 2c Algorithm Phase Two Failure Management USER INTERFACE REQUIREMENTS Tuning DAQ Additions and Modifications Tuning SWARM Global Parameters Other Parameters Control DATABASE REQUIREMENTS Phase One Failure Management Model Global Parameters Detector Parameters SWARM Database Requirements Freeway Structures Page 1 of 47

3 7.2.2 Other Requirements Global Tables DATA DICTIONARY Page 2 of 47

4 LIST OF FIGURES Figure 1-1 System Wide Adaptive RMS Flow... 4 Figure 3-1 Data Flow Diagram... 8 Figure 4-1 Detector Status State Diagram... 9 Figure 5-1 SWARM 1 Forecasting Theory LIST OF TABLES Table 4-1 Detector Status States Table 5-1 Default and Minimum Rate Control Table 7-1 Global Parameters Table 7-2 Typical Values for Volume Threshold Test Table 7-3 Additional Detector Data Table 8-1 Data Dictionary Structure Page 3 of 47

5 1.0 INTRODUCTION 1.1 Purpose of Document Dataflow drawings Detailed description of implementation methodology Database requirements User interface requirements Pseudo code Detailed data dictionary Raw Data Minimum Metering Rates Time of Day Metering Rates Failure Mgmt. Default Rates Start-Up Shut-Down Stategies Data Normalization Station Statistics Incident and Special Event Response Plans Queueing Override Strategies CONTROLS Mode Selection Operational Tuning Parameters Dynamic Saturation Analysis System Wide Adaptive RMS Algorithm Metering Rates Figure 1-1 System Wide Adaptive RMS Flow Figure 1-1 represents a generalized overview of the SWARM concept. While this document is primarily a detailed design of the SWARM system, it includes a design of several necessary components that are required to successfully implement ramp metering. This includes a definition of the basic data required, a first order failure management model, and data normalization. Page 4 of 47

6 1.2 Document Organization The remainder of this document is organized as follows: Chapter 2 provides basic information regarding the operation of the existing ramp metering system. The High-Level Software Architecture is provided in Chapter 3. Chapter 4 provides the Detailed Design for DAQ Modifications and Additions, and Chapter 5 discusses RMS Modules (includes pseudo code). User interface requirements are discussed in Chapter 6. Chapter 7 provides database requirements and Chapter 8 provides the data dictionary. Page 5 of 47

7 2.0 BASIC INFORMATION 2.1 Detector Data The detector data from the controller consists of mainline volume, speed, and occupancy and ramp volume provided from each detector (or detector pair) every 20 seconds. This data will be passed through a first level failure management process then processed to provide the following statistics by lane and by station: Volume Raw Occupancy (un-normalized) Measured speed Normalized Occupancy (removing the effective loop length from the computation) Computed estimated speed (if measured is not available) Density (using available speed and volume) Only normalized statistics will be used in the SWARM algorithms. In addition, normalized saturation density will be computed for each mainline VDS in the system. 2.2 System Configuration Several definitions of the system are required by the SWARM algorithms. The following describes the primary configurations: Limits- Defines the system by the endpoints of each directional freeway. Links-Defines the system by major directional portions of the freeway with differing operational features. This configuration is used to define differing vehicle length and free speed assignments. Break points occur at all freeway-to-freeway interchanges and locations featuring lane drops and additions. Sections-Defines the freeway by directional portions featuring a bottleneck at the downstream endpoint. It is also the key definition for SWARM 1 computations and metering apportionment. Page 6 of 47

8 3.0 HIGH LEVEL SOFTWARE ARCHITECTURE This document contains the detailed design of the SWARM system and the necessary modifications and additions to the existing Data Acquisition (referred to in this document as DAQ) modules required by the design. Figure 3-1 represents the overall dataflow for the SWARM algorithms and the DAQ and FEP module. 3.1 Data Acquisition (DAQ) Modifications and Additions The following modules provide additions and modifications to the DAQ module: Phase One Failure Management Provides global Detector Failure Management Data Normalization and Basic Statistics Provides Normalization of Occupancy, Speed and Density. 3.2 RMS Modules The following modules represent the Ramp Metering System including the SWARM algorithms (SWARM 1 is the network-wide algorithm): Mode Manager This executive module provides control over the RMS based on the user supplied configuration details. Dynamic Saturation Density Computation This module computes the saturation densities for every mainline VDS in the system. SWARM 1 Forecaster This module computes the required values of density at all bottleneck locations based on a linear forecast. The value of density at a time in the future is forecasted and is tested to see if it will exceed saturation density. The required density is computed based on the amount that the forecast predicts exceeding saturation. SWARM 1 Metering Rate Apportionment This module determines the metering rates for each metered ramp in the system. It uses the required densities computed by the linear forecaster, SWARM 2 rates, and user controls to determine the desired rates at each ramp. When ramps cannot accommodate the desired rate excesses are propagated upstream. SWARM 2c (Regional Traffic Responsive) Algorithm This algorithm computes metering rates based on local and downstream conditions. The algorithm attempts to improve on those based on simply local conditions. Page 7 of 47

9 RMS DATAFLOW (VERSION 2.0) HISTORICAL DATA USER INPUT TUNING PARAMETER BOTTLENECK DEFINITION SWARM I REQUIRED DENSITY APPORTIONMENT AND METERING RATE CALCULATOR STATION DENSITY METERING RATES FROM SWARM II TUNING PARAMETERS SWARM II STATION DENSITY DB STATION DENSITY STATION SPEED STATION VOLUME BASIC DATA DATA STATUS FAILURE MANAGEMENT T.O.D. RATES/TABLE MANUAL RATES DB MANUAL RATES MESSAGE MANUAL RATES INITIALIZATION SETTINGS MODE CHANGE RAMP METERING MODE MANAGER T.O.D RATES TABLE SWARMS METERING RATE SWARMS METERING RATE QUEUE OVERRIDE CONTROLLER OVERRIDE INPUT ENABLE MODE RATE METERING ENABLE LIVE DATA (OCC., VOL.) DAQ TICK, VOLUME FEP INTERFACE Figure 3-1 Data Flow Diagram Page 8 of 47

10 4.0 DAQ MODIFICATIONS AND ADDITIONS DETAILED DESIGN 4.1 Phase One Failure Management Model The addition of a Phase One Failure Management module to DAQ s represents a process to determine the status of each loop in the system. This is a required prerequisite to all ramp metering functions since many decisions are based on the validly of the data. The Phase One Failure Management model described below provides checks for volume, occupancy, and the ratio of volume to occupancy for a range of automatically applied conditions Description Phase one failure management is a global model to be applied prior to any analysis or use of the detector data. The following state diagram and table describe the six possible statuses of a detector: Software INHIBITED DISABLED (operator can Disable from any State) HARD FAILED Software Operator or Software Software OK Fail Thres y Times Software Fail Thresh Test Pass Thresh Test Pass Thresh z Times SUSPECT Fail Thresh x Times SOFT FAILED Figure 4-1 Detector Status State Diagram Page 9 of 47

11 Table 4-1 Detector Status States STATUS DATA DESCRIPTION Disabled None The detector has been disabled by the Operator Inhibited None A higher level failure (controller, comm ) is inhibiting the detector form providing data Hard Failed None The detector has either been Soft Failed for a set number of consecutive periods or has failed the threshold test a set number of times during the day. Hard Failed detectors will automatically recover but only after a significant number of consecutive good data conditions. Typically this is set to what amounts to 15 minutes. Soft Failed None The detector has failed the threshold test a predetermined number of consecutive periods. A Soft Failed detector must pass the threshold test a predetermined consecutive number of times to automatically be deemed OK. Suspect OK The detector has failed the threshold test less than a predetermined number of consecutive periods and is deemed suspect. Data is considered good from suspect detectors. OK OK The detector has passed all of the threshold tests. Phase One Failure Management proceeds as follows: For each freeway link 1. If there are any declared incidents or roadwork in the freeway link do not perform failure management on any detectors in that section. 2. Calculate the link average volume, occupancy, and volume/occupancy by detector class as follows: c Detector class mainline, HOV, ramp, collector/connector Nc Number of good detectors of class c in the section Vi,c Volume for detector I, class c Oi,c Occupancy for detector I, class c IF Nc = 0 THEN Vc and Oc are set to a placeholder value (-9999) ELSE Compute Vc = Σ Good Vi,c / Nc Average volume for "good" data in the link by class c Oc = Σ Good Oi,c / Nc Average occupancy for "good" data in the link by class c The central feature of the model consists of testing the data from each detector against a set of high and low thresholds (threshold test). The thresholds are dynamic based upon current freeway conditions. Freeway conditions are divided into freeway links and detector classes (mainline, demand, ramp, etc.). Page 10 of 47

12 3. Then for each detector in the link perform the following: Based on the detector's class (mainline, ramp, etc.) uses the values of Vc and Oc, computed above to find the appropriate set of thresholds and values of Suspect_to_Soft and Soft_to_Ok. Then for each detector in the class compare its actual values of volume, and occupancy against the appropriate high and low thresholds (Note: for detectors not providing occupancy data, only the volume test is performed). To pass the threshold test a detector must pass all tests (high and low for; volume and occupancy (occupancy where applicable)) Inputs and Outputs Inputs: Volume from all detectors and Occupancy from Mainline detectors Outputs: Status for all detectors Pseudo Code /Compute link average volumes and occupancy/ /clear all summators/ FOR link=1 TO number_links FOR type=1 TO number_types avg_link_vol(link,type)=0 avg_link_occ(link,type)=0 good(link,type)=0 total(link,type)=0 NEXT type,link /sum good volumes and tick counts by loop type and freeway link/ FOR i=1 TO total_number_loops GET{link_ID, loop_type,loop_status} FROM TABLE {loop_config}for {loop_id=i} total(link_id,loop_type)=total(link_id,loop_type)+1 IF{loop_status= ok or loop_status = suspect } THEN /get current lane_volume and occupancy for loop i avg_link_vol(link_id,loop_type)=avg_link_vol(link_id,loop_type)+lane_volume(i) avg_link_occ(link_id,loop_type)=avg_link_occ(link_id,loop_type)+occupancy(i) good(link_id,loop_type)=good(link_id,loop_type)+1 NEXT I /compute average volumes, occupancies and volume/occupancy for each link and loop type/ FOR link=1 TO number_links FOR type=1 TO number_types IF {good(link,type)/total(link,type)>0} THEN avg_link_vol(link,type)=avg_link_vol(link,type)/good(link,type) avg_link_occ(link,type)=avg_link_occ(link,type)/good(link,type) ELSE avg_link_vol(link,type)= avg_link_occ(link,type)= NEXT type,link Page 11 of 47

13 /perform Phase I Failure Management FOR i=1 TO total_number_loops GET{link_ID, loop_type,loop_status} FROM TABLE {loop_config}for {loop_id=i} SELECT CASE {loop_status(i)} CASE 1 /previous status was ok / CALL ThreshTest(i,link, type, status) IF {status =0} THEN /all threshold tests were passed/ pass_fail_counter(i) = 0 ELSE pass_fail_counter(i)=pass_fail_counter(i)+1 suspect_counter(i)=suspect_counter(i)+1 loop_status(i)=2 / suspect / IF {suspect_counter(i) >=max_fail_to_hard} THEN loop_status(i) = 4 /"hard failed"/ pass_fail_counter(i) = 0 soft_to_hard_counter(i) = 0 CASE 2 / previous status was "suspect"/ CALL ThreshTest(i,link, type, status) IF {status =0} THEN /all threshold tests were passed/ pass_fail_counter(i) = 0 loop_status(i)=1 /"ok"/ ELSE suspect_counter(i)=suspect_counter(i)+1 IF {suspect_counter(i) >=max_fail_to_hard} THEN loop_status(i)=4 /"hard failed"/ pass_fail_counter(i) = 0 soft_to_hard_counter(i) = 0 pass_fail_counter=pass_fail_counter+1 IF {Pass_Fail_Counter >= Current Value of suspect_to_soft} THEN loop_status(i)=3 /"soft failed" / pass_fail_counter(i) = 0 CASE 3 /"soft failed"/ CALL ThreshTest(i,link, type, status) IF {status <>0} THEN /at least one threshold test was not passed/ pass_fail_counter(i) = 0 soft_to_hard_counter(i)=soft_to_hard_counter(i)+1 IF {soft_to_hard_counter(i) >= soft_to_hard} THEN loop_status(i)=4 /"hard failed"/ soft_to_hard_counter(i) = 0 ELSE /all threshold tests were passed/ pass_fail_counter(i)=pass_fail_counter(i)+1 IF {pass_fail_counter(i) > = Current Value of soft_to_ok} FROM Occupancy_threshold_Table] THEN loop_status(i)=1 / "ok pass_fail_counter(i) = 0 Page 12 of 47

14 soft_to_hard_counter(i) = 0 ELSE suspect_counter(i)=suspect_counter(i)+1 IF {suspect_counter(i) > = max_fail_to_hard} THEN loop_status(i) = 4 / "hard failed" / pass_fail_counter(i) = 0 soft_to_hard_counter(i) = 0 CASE 4 /"hard failed"/ CALL ThreshTest(i,link, type, status) IF {status <>0} THEN /at least one threshold test was not passed/ pass_fail_counter(i) = 0 ELSE /all threshold tests were passed/ pass_fail_counter(i)=pass_fail_counter(i)+1 IF {pass_fail_counter(i) > = Global Hard to Ok Parameter} THEN loop_status(i)=1 / "ok pass_fail_counter(i) = 0 ELSE END SELECT NEXT I SUBROUTINE ThreshTest (link, type, status) status=0 GET {thresh_low_vol, thresh_high_vol, suspect_to_soft, soft_to_ok} FROM TABLE Volume_Thresholds FOR CONDITION {Actual_vol=average_vol(link, type)) OR IF average_vol(link,type) >=0 THEN GET {thresh_low_vol, thresh_high_vol, suspect_to_soft, soft_to_ok} FROM TABLE Volume_Thresholds FOR CONDITION {Actual_vol= default} IF {lane_volume(i)<thresh_low_vol OR lane_volume(i) > thresh_high_vol} THEN status=status+1 IF type = mainline THEN GET {thresh_low_occ, thresh_high_occ,} FROM TABLE Occupancy_Thresholds FOR CONDITION {Actual_occ=average_occ(link, type)) OR if Occ(link,type) < 0 THEN GET {thresh_low_occ, thresh_high_occ,} FROM TABLE Occupancy_Thresholds FOR CONDITION {Actual_occ=default) IF {Occupancy(i) < thresh_low_occ OR tick_count(i)/9 > thresh_high_occ} THEN status=status+1 RETURN Page 13 of 47

15 Once per Day For all detectors suspect_counter(i) = 0 Operator Intervention When the Operator intervenes to repair or enable a detector, its state is forced to "ok". IF {Operator Disables Detector i} THEN loop_status(i) = 7 /"disabled"/ pass_fail_counter(i) = 0 soft_to_hard_counter(i) = 0 suspect_counter(i) = 0 ELSEIF {Operator Repairs or Enables Detector i} THEN loop_status(i) = 1 /"ok"/ pass_fail_counter(i) = 0 soft_to_hard_counter(i) = 0 suspect_counter(i) = 0 Inhibiting Detectors /Note: Disabled detector cannot be inhibited/ IF{Detector i is being inhibited AND is not currently "disabled"} THEN previous_status(i)=loop_status(i) loop_status(i) = 5 /"inhibited"/ ELSEIF {Detectors are being uninhibited} THEN loop_status(i) = previous_status(i) 4.2 Data Normalization and Basic Statistics Description Occupancy is normalized by determining the effective loop length for each VDS loop. This is accomplished by computing the effective loop length from the estimated speed equation. When measured speed is available the computation continues at all times of day. If only estimated speed is available, an assumed free speed during certain fixed periods of the day are used to compute the effective loop length. The computation uses a slow exponential filter, guaranteeing a large sample size. The computed effective loop length is removed from the occupancy value producing a normalized statistic. The final result is to provide normalized and unnormalize values of occupancy, speed, and density for each lane and VDS. k, l 52.8tS k, lo = 3600V k, l k, l λ when the speed is measured l IF H e1 < k, l < H e2 THEN OR for RDMS data IF RH e1 < k, l < RH e2 THEN e = P + e (1 P k, l k, l ek k, l ek ) Normalized occupancy is calculated using the previously calculated effective loop length in this equation: Page 14 of 47

16 θ k, l λl 100* Ok, l λl + ek, l = λl 100 Ok, l *(1 λ + e l k, l ) Estimated Speed S k,l = 3600 Vk, l 52.8 t θ k, l λl (Miles per Hour) Inputs and Outputs INPUT VARIABLES {loop_status(k,l), tick_count(k,l), lane_volume(k,l), ell(k,l)} INPUT PARAMETERS {polling_interval, scan_rate, average_vehicle_length(j,l), ell_max_volume, ell_min_speed, current_tod, ell_tod_start, ell_tod_stop, free_speed(j,l), ell_min_length, ell_max_length, ell_smoothing_parameter, number_physical_lanes(k)} OUTPUT VARIABLES {lane_occupancy(k,l), lane_speed(k,l), lane_density(k,l), station_volume(k), station_occupancy(k), station_speed(k), station_density(k),norm_lane_occupancy(k,l), norm_lane_speed(k,l), norm_lane_density(k,l), station_volume(k), norm_station_occupancy(k), norm_station_speed(k), norm_station_density(k) station_status(k), ell(k,l)} Pseudo Code FOR {each freeway link j} FOR {each mainline VDS k in link j} sum_good_lanes=0 sum_lane_volume=0 sum_lane_occupancy=0 sum_lane_speed=0 sum_norm_lane_occupancy=0 sum_norm_lane_speed=0 FOR {each lane l at VDS k} IF {measured speed is available} THEN ell_temp=(52.8 free_speed(j,l) lane_occupancy(k,l))/lane_volume(k,l)-average_vehicle_length (k,l) IF {ell_temp >ell_min_length AND ell_temp < ell_max_length} THEN ell(k,l)=ell_smoothing_parameter ell_temp+(1-ell_smoothing_parameter) ell(k,l) ELSE IF {loop_status(k,l) <= 2} THEN ; status is ok or suspect lane_occupancy(k,l)=100 tick_count(k,l)/(polling_interval scan_rate) lane_speed(k,l)=((average_vehicle_length(j,l)+ell(k,l)) lane_volume(k,l) 3600/polling_interval) /(52.8 lane_occupancy(k,l)) IF {lane_volume(k,l)<ell_max_volume AND lane_volume(k,l) > 0 AND lane_speed(k,l) > ell_min_speed AND current_tod > ell_tod_start AND current_tod < ell_tod_stop} THEN Page 15 of 47

17 ell_temp=(52.8 free_speed(j,l) lane_occupancy(k,l))/lane_volume(k,l)-average_vehicle_length (k,l) IF {ell_temp >ell_min_length AND ell_temp < ell_max_length} THEN ell(k,l)=ell_smoothing_parameter ell_temp+(1-ell_smoothing_parameter) ell(k,l) sum_lane_volume=sum_lane_volume+lane_volume(k,l) sum_lane_occupancy=sum_lane_occupancy+lane_occupancy(k,l) norm_lane_occupancy(k,l) θ k, l λl 100* Ok, l λl + ek, l = λl 100 Ok, l *(1 λ + e sum_norm_lane_occupancy=sum_lane_occupancy+norm_lane_occupancy(k,l) IF {lane_occupancy(k,l) > 0} THEN IFlane_speed(k,l) I s not measured THEN lane_speed(k,l) average_vehicle_length(j,l) lane_volume(k,l) 3600/polling_interval /(52.8 lane_occupancy(k,l)) sum_lane_speed=sum_lane_speed+lane_speed(k,l) lane_volume(k,l) norm_lane_speed(k,l)=average_vehicle_length(j,l) lane_volume(k,l) 3600/polling_interval /(52.8 norm_lane_occupancy(k,l)) sum_norm_lane_speed=sum_norm_lane_speed+norm_lane_speed(k,l) lane_volume(k,l) lane_density(k,l)=lane_volume(k,l) 3600/polling_interval/lane_speed(k,l) norm_lane_density(k,l)=lane_volume(k,l) 3600/polling_interval/norm_lane_speed(k,l) ELSE lane_speed(k,l)=0 norm_lane_speed(k,l)=0 lane_density(k,l)=0 norm_lane_density(k,l)=0 number_good_lanes=number_good_lanes+1 NEXT {lane} IF {number_good_lanes>0} THEN station_volume(k)=3600 number_physical_lanes(k) sum_lane_volume /(number_good_lanes polling_interval) station_occupancy(k)=sum_lane_occupancy/number_good_lanes norm_station_occupancy(k)=sum_norm_lane_occupancy/number_good_lanes station_speed(k)=sum_lane_speed/(sum_lane_volume 3600/polling_interval) norm_station_speed(k)=sum_norm_lane_speed/(sum_lane_volume 3600/polling_interval) station_density(k)=station_volume(k)/station_speed(k)/number_physical_lanes(k) norm_station_density(k)=station_volume(k)/norm_station_speed(k)/number_physical_lanes(k) station_status(k)=number_good_lanes/number_physical_lanes(k) ELSE station_volume(k)= 0 station_occupancy(k)= 0 norm_station_occupancy(k)= 0 station_speed(k)= 0 norm_station_speed(k)= 0 Page 16 of 47 l k, l )

18 station_density(k)= 0 norm_station_density(k)= 0 station_status(k)=0.0 NEXT {VDS} NEXT {freeway link} Page 17 of 47

19 5.0 RMS MODULES 5.1 Mode Manager Description The RMS Mode Manager is the executive task responsible for determining the current mode of RMS operation, queuing up the appropriate central algorithms, distributing the metering rates, and startup/shutdown. The modes are defined in the following table: MODE DESCRIPTION Disabled Metering Disabled at this ramp (Signal Off) Local TR Local Traffic Responsive Operation Local TOD Local Time of Day Operation The following SWARM modes are available: SWARM 2c SWARM 2c Operation only SWARM 1 SWARM 1 Operation only SWARM 1 and 2c SWARM 1 and SWARM 2c Operation In addition to selecting the mode of operation certain SWARM controls are required for setting default and minimum metering rates and system startup method. The following describes these three controls: Default Minimum Rates During SWARM metering rate computations, either TOD rates (Rtod) or the absolute minimum rate (Rmin) is used as the minimum metering rates depending on this operator selectable parameter. Default Metering Rates When SWARM is chosen for operation, and cannot operated due to equipment failures, either TOD or local maximum (Rmax) rates will be used as the default metering rate depending on this operator selectable parameter. These two parameters yield the following configurations: Table 5-1 Default and Minimum Rate Control Operator Selected Minimum Rate Operator Selected Default Rate Resulting SWARM Minimum Rate Resulting SWARM Default Rate TOD TOD Rtod Rtod TOD ABS MAX Rtod Rmax ABS MIN TOD Rmin Rtod ABS MIN ABS MAX Rmin Rmax Page 18 of 47

20 Metering Operation For metering to be started or shutdown SWARM must be directed as to the method. Two choices are provided; strictly based on TOD tables, or both on TOD tables and freeway conditions. The freeway conditions methodology is as follows. Startup If a user definable number of consecutive 20 second periods requiring less than maximum rates has been determined metering operations are started. Shutdown If a user definable number of consecutive 20 second periods requiring maximum rates has been determined metering operations are shutdown. 5.2 Dynamic Saturation Density Description This module computes the saturation density at all VDS locations. This is accomplished by fitting a parabola to data near saturation conditions. The module runs continuously and uses relative large data samples to fit the parabola. When a set number of data points have been collected, the coefficients of a parabola forced through the origin are computed and the maximum value determined (saturation density). This value is then exponentially filtered with the old value to provide a dynamically determined value of saturation density Inputs and Outputs INPUT VARIABLES {station_status(k), station_volume(k), norm_station_density(k), saturation_density(k)} STATIC VARIABLES { sum_d(k)=[0], sum_d2(k)=[0], sum_d3(k)=[0], sum_d4(k)=[0], sum_v(k)=[0], sum_vd(k)=[0], sum_vd2(k)=[0], number_sat_density(k)=[0]} INPUT PARAMETERS {percent_good_lanes, number_physical_lanes(k), max_volume, sample_size_sat_density, sat_smoothing_parameter} OUTPUT VARIABLES {saturation_density(k)} Page 19 of 47

21 5.2.3 Pseudo Code FOR {each VDS k} IF {station_status(k) 100 >percent_good_lanes} THEN IF {station_volume(k)>number_physical_lanes(k) max_volume/2} THEN sum_d(k)=sum_d(k)+norm_station_density(k) sum_d2(k)=sum_d2(k)+norm_station_density(k) 2 sum_d3(k)=sum_d3(k)+norm_station_density(k) 3 sum_d4(k)=sum_d4(k)+norm_station_density(k) 4 sum_v(k)=sum_v(k)+station_volume(k) sum_vd(k)=sum_vd(k)+station_volume(k) norm_station_density(k) sum_vd2(k)=sum_vd2(k)+station_volume(k) norm_station_density(k) 2 number_sat_density(k)=number_sat_density(k)+1 IF {number_sat_density(k) >= sample_size_sat_density} THEN dbar=sum_d(k)/number_sat_density(k) vbar=sum_v(k)/number_sat_density(k) dbar2=sum_d(k)/number_sat_density(k) s1v=sum_vd(k)-number_sat_density(k) vbar dbar s2v=sum_vd2(k)-number_sat_density(k) dbar2 vbar s11=sum_d2(k)-number_sat_density(k) dbar dbar s12=sum_d3(k)-number_sat_density(k) dbar dbar2 s22=sum_d4(k)-number_sat_density(k) dbar2 dbar2 b2=(s2v s11-s1v s12)/( s11 s22-s12 s12) b1=(s1v s22-s2v s12)/(s11 s22-s12 s12) saturation_density(k)=sat_smoothing_parameter saturation_density(k) +(1-sat_smoothing_parameter) -b1/2/b2 sum_d(k)=0 sum_d2(k)=0 sum_d3(k)=0 sum_d4(k)=0 sum_v(k)=0 sum_vd(k)=0 sum_vd2(k)=0 number_sat_density(k)=0 NEXT {VDS} 5.3 SWARM 1 Algorithm Description The following describes the SWARM 1 methodology: Consider the freeway network being divided into sections. Each section has as its furthest downstream point a known bottleneck. By definition a bottleneck represents a location whose geometrics causes it to be the most restrictive location in the section. Therefore, the SWARM 1 algorithm operates at bottleneck locations and controls all upstream locations in the section relative to the flow at the bottleneck. Page 20 of 47

22 Density will be the control parameter for the SWARM 1 algorithm. Since density is directly related to congestion, it will be monitored at each bottleneck location. The algorithm requires a nominal saturation density threshold value for each VDS in the freeway network. Saturation density levels are updated in real time for each operational VDS based on data near expected saturation values. This is accomplished by fitting a second-degree polynomial to the volume/density data then obtaining the maximum value of density. This represents saturation density and is exponentially smoothed with the previous value using relatively slow parameters. Based upon recent conditions, the algorithm will attempt to estimate the density at a time in the future. How far into the future (T crit ) the algorithm will estimate will be a tunable parameter and will be based upon practical limitations of the algorithm and the necessary lead time for metering rates to take effect. If the estimated density exceeds the bottleneck s saturation density then ramp metering rates will be computed in an attempt to head off the predicted onset of congestion. Figure 5-1 depicts this forecasting methodology. The heavy line on the main drawing represents density rising in a typical fashion. In the Figure the forecast indicates that density will rise above saturation before T crit minutes. The heavy line above the saturation density line represents the amount that the density must be reduced during the next T crit minutes to avoid saturation. Every 30 seconds a new forecast is made and if saturation density will occur before the T crit minutes, a new density reduction is computed. The theory behind the forecast is as follows: Forcasted Density SATURATION DENSITY Reduction in Density Req. During Next Tcrit Time Periods DENSITY Tcrit Actual Density Corrective Trend at time t TIME Figure 5-1 SWARM 1 Forecasting Theory The current value of density must be less than saturation density to proceed with a forecast. If the current density is greater than or equal to saturation density, the forecasting portion of the algorithm is skipped. A density trend will be estimated based on a tunable set of consecutive density values obtained at the bottleneck VDSs. The size of the data set is a tunable parameter. Forecast model: Assume the current time is t. Then the basic forecast model is that density at time t+1 equals the density at time t plus b, where b represents the forecasted slope of the density trend at time t. The key to the forecast is to obtain the best possible estimate of b. To accomplish this objective two steps are required; first estimate the slope of the line in the interval (t-h) to t using a simple linear curve technique. Using the slope of the equation compute the density at time Tcrit. Page 21 of 47

23 If the forecasted value of density exceeds the bottleneck's saturation density then the required bottleneck density will be computed as follows as follows: current density minus 1/Tcrit times the amount of excess density (above saturation density) computed above. If the forecasted density is less than saturation density or the current density is greater than or equal to saturation density then the required bottleneck density equals saturation density SWARM 1 Metering Rate Apportionment and Propagation The process for computing SWARM 1 metering rates is a two-pass process; the first pass determines rates and computes excesses for propagation only considering contiguous freeways, the second pass recomputes the rates including the propagated excesses for non-contiguous freeways. Based on the required bottleneck densities computed by the SWARM 1 algorithm, metering rates are computed as follows in both Pass 1 and Pass 2. For each section defined in the Dynamic Section Table: For every VDS immediately upstream of each metered ramp, first compute the reduction in volume (VR) required to reduce the local density at that ramp. This is equal to the (local density - required bottleneck density) times (number of lanes) times (distance to the next downstream ramp). Then starting at the bottleneck and working upstream compute metering rates based on current ramp volumes and the required volume reductions. Actual metering rates will be subject to SWARM 2 rates and maximum and minimum rates. Since reductions may be positive or negative, excess or surplus values are propagated upstream within each section. Intra-section propagation proceeds as follows: Pass 1 If at the end of a freeway section, positive excess still remains; it is propagated to the next upstream section(s) according to the following: If the section is connected to only a contiguous freeway section, then propagate the required amount to the next section. If the section is connected to multiple sections, then propagate the appropriate amount to the contiguous section and record the amounts for the non-contiguous sections. Pass 2 For each section test if a propagated amount was recorded during Pass 1. If not proceed to the next section, if yes, apply the excess and recompute the metering rates in the section and all subsequent upstream sections. Page 22 of 47

24 5.3.3 Pseudo Code LINEAR FORECASTER INPUT VARIABLES {norm_station_density(i,t), saturation_density(i)} INPUT PARAMETERS {sat_density_multiplier(i), forecast_lead_time, density_sample_size} t={current time} FOR each bottleneck VDS i from TABLE test_good_data=0 DO IF {station_status < percent_good_lanes} THEN /bottleneck VDS is failed/ test_good_data=test_good_data+1 GET NEXT VDS IF{ test_good_data > number_vds_oper} THEN EXIT DO WHILE test_good_data>0 IF {test_good_data <= number_vds_oper} THEN IF{ norm_station_density(i,t) > saturation_density(i)} THEN required_density(i) = saturation_density(i) _sat_density_multiplier(i) ELSE sumt=0 sumd=0 sumtd=0 sumt2=0 FOR j= t-density_sample_size to t sumt = sumt + j sumd = sumd +norm_station_density(j) sumtd = sumtd+ j norm_station_density(j) sumt2 = sumt2 j*j NEXT j b= (sumtd-sumt*sumd/ density_sample_size)/(sumt2-sumt*sumt/ density_sample_size) a=sumd/ density_sample_size-b*sumt/ density_sample_size forecast_station_density=a+b*(t+ forecast_lead_time) IF{ forecast_station_density < saturation_density(i) sat_density_multiplier(i)} THEN required_density(i) = forecast_station_density ELSE required_density(i) = saturation_density(i) sat_density_multiplier(i) NEXT bottleneck VDS i OUPUT VARIABLES {required_density(i), forecast_station_density(i)} Page 23 of 47

25 Apportionment and SWARM 1 Rates {pass 1} excess=0 FOR k= 1 TO {number of freeway sections} Mile_post=bottleneck_milepost FOR j=1 TO {number of entrance ramps in the section} required_volume_reduction=avs(ramp_mile_post)*(norm_station_density{upsteream VDS} required_density) *number_physical_lanes {upstream VDS} mile_post=ramp_mile_post desired_metering_rate=ramp_volume-requried_voulume_reduction-excess IF {ramp is metered} sw1_meter_rate=max(minimum_meter_rate, MIN(maximum_meter_rate, sw2_meter_rate)) ELSE sw1_meter_rate=ramp volume Excess=section_prop_factor*(sw1_meter_rate-desired_metering _rate) IF {more than one upstream ramp AND Excess>0} excess_non_contigious_1=excess*intra-section_prop_factor*excess_non_contigious_factor_1 excess_non_contigious_2=excess*intra_section_prop_factor*excess_non_contigious_factor_2 Excess=Excess excess_non_contigious_1 excess_non_contigous2 ELSE excess_non_contigious_1=0 excess_non_contigious_2=0 NEXT {ramp} Excess=intra_section_prop_factor*Excess NEXT {section} {pass 2} Excess=0 FOR k = 1 TO {number for freeway sections} Mile_post=bottleneck_milepost FOR j = 1 TO {number of entrance ramps in the section} Excess=Excess+excess_non_contigious_2 required_volume_reduction=avs(ramp_mile_post)*(norm_station_density{upstream VDS}- requred_density)*number_physical_lanes{upstream VDS} mile_post=ramp_mile_post desired_metering_rate=ramp_volume-required_volume_reduction-excess IF {ramp is metered} sw1_meter_rate+max(minimum_meter_rate, MIN(maximum_meter_rate, sw2_meter_rate)) ELSE sw1_meter_rate=ramp volume Excess=section_prop_factor*(sw1_meter_rate-desired_metering_rate) NEXT {ramp} Excess=intra_section_prop_factor*Excess NEXT{section} Page 24 of 47

26 5.4 SWARM 2c Algorithm Concept This algorithm uses data from two mainline sites, the one adjacent to the metered ramp and the one downstream. The occupancy from the two sites is averaged and smoothed and used to lookup the metering rate from a local table. The rate is then smoothed and sent to the Mode Manager to be sent to meters when the LTR mode is chosen. The algorithm operates continuously but is only used based on Controls and Mode Selection. The advantage of this algorithm is that rates are based on an area view of traffic. Controls and Parameters Minimum Rate Control This control affords the user the ability to have LTR choose a minimum rate from either the Time of Day (TOD) table or the Absolute Minimum rate. Times of Operation Operation can be either during the same periods as the TOD tables or between two times of day (M-F) Turn On/Turn Off Control This set of parameters determine if metering is needed. If the raw metering rate is at or above the Absolute Maximum Rate for the number of polling cycles given by these two parameters, metering is either turned on or off. Minimum Acceptable Mainline Data This global parameter defines the percentage of good data from a mainline station to be deemed operational Mainline Smoothing Parameter A value between 0.10 and 0.99 representing the ratio of current mainline Occupancy to be smoothed with the last smoothed value Metering Rate Smoothing Parameter A value between 0.10 and 0.99 representing the ratio of the latest rate from the rate table to be smoothed with the last smoothed value. Metering Rate Table A table correlating Occupancy to Metering Rate. Values from the Absolute Minimum Rate to the Absolute Maximum Rate can be obtained. Operation At the end of each polling cycle (20 60 seconds) the algorithm follows the following steps: 1. Determine if the Mainline Data is ok. The local mainline station must be operational or a message is sent to the Mode Manager indicating that a backup mode must be used. If only the local station is operational rates are determined from that station alone. 2. The raw mainline data is smoothed and a rate is looked up in a table. 3. The rate is then smoothed with the old rate and bounded by the controls shown above. 4. Special rates of 0 (turn metering off) and -1 (cannot determine rate) are also part of the algorithm. Mode Manager The Mode Manager is the interface between the Operator and the ramp metering software. It allows the Operator to select from the possible modes (Disable, Fixed, TOD, LTR) and determines the final metering rates sent to the controllers. It indicates various information to the Operator as well as Desired Mode and Actual Mode. If LTR sends a -1 rate it will switch to TOD mode as the actual mode. Page 25 of 47

27 The following is the Pseudo Code for the LTR Algorithm. ; Inputs ; LocalMainlineOccupancy Occupancy at the mainline adjacent to the metered ramp ; DownstreamMainlineOccupancy Occupancy at the next downstream mainline site ; LocalMainlineStatus Status ( 1.0) of local mainline station ; DownstreamMainlineStatus Status ( 1.0) of downstream mainline station ; MLSmoothingFactor Global Tunable Parameter for smoothing mainline occupancy (0.10 to 0.99 default 0.50) ; LTRpercentGoodData Global Tunable Parameter ( 1.0 default = 0.19) indicating required mainline status (not actually a percentage) ; Rate Lookup Table A table for each ramp matching Mainline Occupancy with a Metering Rate (Rates go from 3 to 18 vehicles per minute with corresponding Occupancy) ; RateSmoothingFactor Global Tunable Parameter for smoothing metering rates (0.10 to 0.99 default 0.50) ; MinRateControl Tunable Parameter indicating for each ramp where the minimum rate is obtained from (either the current TOD rate or Absolute Minimum rate) ; AbsoluteMinRate - Tunable Parameter indicating for each ramp the Absolute Minimum metering rate ( 3 to 10 vehicles per minute default 4) ; AbsoluteMaxRate - Tunable Parameter indicating for each ramp the Absolute Maximum metering rate (10 to 18 vehicles per minute default 15) ; CurrentTODrate The current metering rate from the associated metered ramps Time of Day Table ; MeterDuring - Tunable Parameter indicating for each ramp when metering is allowed (either when the TOD tables indicate metering or between specified times) ; MaxONtoOFF Tunable Parameter indicating for each ramp how many consecutive polls where the RawRate is greater than or equal to the AbsoluteMaxRate when the meter is currently on. Once this threshold is met the meter is turned off (internally) ; MaxOFFtoON - Tunable Parameter indicating for each ramp how many consecutive polls where the RawRate is less than the AbsoluteMaxRate when the meter is currently off. Once this threshold is met the meter is turned on (internally) ; ; ; Internal Variables ; for each ramp ; MLRawOccupancy The raw value of the average occupancy of the good stations ; MinRate the minimum metering rate based on MinRateControl ; LTRon-off Keeps track of whether LTR thinks metering should be on or off ; RawRate - The raw value of the unsmoothed metering rate ; MaxCounter a counter that keeps track of the number of consecutive polls meeting the condition for switching a meter on or off ; ; ; Outputs ; for each ramp ; LTRrate The current desired rate from this algorithm. The rate produced can be any value from AbsoluteMinRate to AbsoluteMaxRate. Also a rate of 0 indicates to turn off the meter and a rate of -1 indicates rate could not be determined (missing data) ; MLSmoothedOccupancy The resultant smoothed occupancy ; SmoothedRate The smoothed rate before being bounded by constraints Page 26 of 47

28 ;Once every polling cycle FOR i = 1 to [number of metered ramps] ; First, determine if the Mainline data meets the minimum standards IF LocalMainlineStatus < LTRpercent good data THEN LTRrate(i) = -1 ; -1 is placeholder indicating MR could not be formed MLRawOccupancy(i) = -1 MLSmoothedOccupancy(i) = -1 RawRate(i) = -1 SmoothedRate(i) = -1 LTRon-off(i) = "Off" MaxCounter(i)=0 RETURN ELSEIF DownstreamMainlineStatus < LTRpercentGoodData THEN MLRawOccupancy(i) = LocalMainlineOccupancy ; use only Local Mainline Occupancy ELSE ; both Mainline Stations have Good Status MLRawOccupancy(i) = (LocalMainlineOccupancy + DownstreamMainlineOccupancy) / 2.0 ; Next Compute the Smoothed Occupancy from the last value of MLSmoothedOccupancy and the new Raw Occupancy IF MLSmoothedOccupancy(i) <= 0 THEN MLSmoothedOccupancy(i) = MLRawOccupancy(i) ; If no previous Smoothed Occ use Raw ELSE MLSmoothedOccupancy(i) = MLRawOccupancy(i) * MLSmoothingFactor +MLSmoothedOccupancy(i) * (1 - MLSmoothingFactor) ; get rate from table RawRate(i) = [lookup in table based on MLSmoothedOccupancy(i)] ; Compute the smoothed rate from the last value of SmoothedRate and the new raw rate IF SmoothedRate(i) <= 0 THEN SmoothedRate(i) =RawRate(i) ; If no previous Smoothed Rate used Raw ELSE SmoothedRate(i)= RawRate(i) * RateSmoothingFactor + SmoothedRate(i) * (1 - RateSmoothingFactor) ; Now Bound the rate IF MinRateControl = "TOD" THEN MinRate= [Current TOD Rate for this Ramp] IF MinRate < AbsoluteMinRate(i) THEN MinRate=AbsoluteMinRate(i) ELSE ; MinRateControl is set to Absolute Min MinRate=AbsoluteMinRate(i) LTRrate(i) = MIN( MAX ( MinRate, SmoothedRate(i)), AbsoluteMaxRate(i) ) ; Now test if LTR should be metering independent of actual desired times Page 27 of 47

29 IF LTRon-off(i) = "On" THEN IF RawRate(i) >= AbsoluteMaxRate(i) THEN MaxCounter(i) = MaxCounter(i) +1 IF MaxCounter(i) >= MaxONtoOFF THEN LTRon-off(i)= "Off" MaxCounter(i) = 0 ELSE MaxCounter(i)=0 ELSE ; LTRon-off(i) = "Off" IF RawRate(i) < AbsoluteMaxRate(i) THEN MaxCounter(i) = MaxCounter(i) +1 IF MaxCounter(i) >= MaxOFFtoON THEN LTRon-off(i) = "On" MaxCounter(i) = 0 ELSE MaxCounter(i) = 0 ; Now test if Metering is allowed IF MeterDuring(i) = "TOD" THEN IF [Current TOD Rate for this Ramp] < AbsoluteMinRate(i) THEN LTRrate(i) = 0 ELSE IF LTRon-off(i) ="Off" THEN LTRrate(i) = 0 ELSEIF [Current Time is between the desired times (inclusive) THEN IF LTRon-off(i) = "Off" THEN LTRrate(I) = 0 ELSE ; not during TOD and outside of desired times LTRrate(i) = 0 NEXT i 5.5 Phase Two Failure Management The primary Phase Two Failure Management issue associated with ramp metering and SWARM is related to procedures for missing data. Two cases must be considered for each VDS station; partial data loss and total data loss. Partial Data Loss If more than a threshold percentage of the lanes associated with a VDS are operational, the station statistics are computed from the good data by a simple averaging technique. If less than threshold percentage of the data is available, the station is considered totally failed. Page 28 of 47

30 Total Data Loss If a VDS is deemed totally failed by the condition stated above, it is removed from the network relative to SWARM operation. For SWARM1, if this is a bottleneck location, the first good upstream VDS from the failed bottleneck VDS is chosen. A settable threshold id associated with each VDS as to how far upstream to search. If no VDS is found to replace a totally failed VDS, SWARM 1 does not operate in that section of freeway. SWARM 2c will not provide a metering rate for a given ramp if the necessary associated mainline VDS is not available or has less than the required percentage of data Page 29 of 47

31 6.0 USER INTERFACE REQUIREMENTS The user interface is provided through three capabilities; monitoring, tuning, and control. Monitoring is provided through reports, screens, and the graphical interface. Tuning is provided through screens, which allow definition of parameters that effect the current operations. Control is provided through screens that allow the capability to configure the current operations. 6.1 Tuning The following parameters are tunable for each process DAQ Additions and Modifications Tuning Phase 1 Failure Management max_fail_to_hard soft_to_hard percent_good_data hard_to_ok Number of times during a day a detector has failed its threshold test to be reclassified as hard failed Number of consecutive minutes a detector in the soft failed status must fail its threshold test to be reclassified hard failed Percentage of good data necessary in each detector class to allow failure management to proceed in a given freeway link Number of consecutive minutes of good data before returning a detector to ok In addition the high and low thresholds for volume, occupancy, and volume/occupancy and the suspect_to_soft and soft_to_ok parameters are tunable according to average values in the dynamic_failure_thresholds table. Effective Loop Length Computation ell_max_length ell_min_length ell_max_vol ell_min_speed ell_smoothing_parameter ell_tod_start ell_tod_stop Absolute maximum loop length Absolute minimum loop length Upper bound on current volume to proceed with computation Lower bound on current speed to proceed with computation Smoothing parameter Start time for computations Stop time for computations SWARM Global Parameters Saturation Density max_volume sample_size_sat_density sat_smoothing_parameter Maximum hourly lane volume Number of samples required to update the VDS s saturation density Smoothing parameter Page 30 of 47

32 Forecaster number_vds_oper forecast_sample_size forecast_lead_time Maximum number of VDS s to search for if the bottleneck VDS is failed Number of time intervals in the past to be used for the forecast Number of time intervals into the future to forecast SWARM 1 Apportionment and Bottlenecks section_prop_factor intra_section_prop_factor bottleneck_flag sat_density_multiplier Percent to propagate within a section Percent to propagate between sections Tells whether a VDS is a bottleneck Amount to multiply the saturation density by to form the final saturation density SWARM 2c MLSmoothingFactor - Global Tunable Parameter for smoothing mainline occupancy (0.10 to 0.99 default 0.50) LTRpercentGoodData Rate Lookup Table Global Tunable Parameter ( 1.0 default = 0.19) indicating required mainline status (not actually a percentage) A table for each ramp matching Mainline Occupancy with a Metering Rate (Rates go from 3 to 18 vehicles per minute with corresponding Occupancy) RateSmoothingFactor Global Tunable Parameter for smoothing metering rates (0.10 to 0.99 default 0.50) MinRateControl MeterDuring - MaxONtoOFF Tunable Parameter indicating for each ramp where the minimum rate is obtained from (either the current TOD rate or Absolute Minimum rate) Tunable Parameter indicating for each ramp when metering is allowed (either when the TOD tables indicate metering or between specified times) Tunable Parameter indicating for each ramp how many consecutive polls where the RawRate is greater than or equal to the AbsoluteMaxRate when the meter is currently on. Once this threshold is met the meter is turned off (internally) MaxOFFtoON - Tunable Parameter indicating for each ramp how many consecutive polls where the RawRate is less than the AbsoluteMaxRate when the meter is currently off. Once this threshold is met the meter is turned on (internally) SWARM Phase 2 Failure Management percent_good_lanes The percentage of good lanes necessary to compute station statistics Page 31 of 47

33 6.1.3 Other Parameters Link Parameters (by lane) average_vehicle_length free_speed Metered Ramp Parameters (by ramp) min_rate max_rate Absolute minimum metering rate Absolute maximum metering rate 6.2 Control Phase 1 Failure Management For each detector loop: loop_status The current status of the loop. Used by the operator to enable and disable a loop. Metered Ramps For each metered ramp: metering_enable swarm_1_enable swarm_2c_enable default_rate_selection minimum_rate_selection swarm_start_stop_strat Represents if metering has been enabled at this ramp Represent if SWARM 1 is enabled in this section Represents if SWARM 2c is enabled at this ramp Represents how default metering rates are selected Represents how minimum rates are determined Represents SWARM start/stop metering strategy (TOD or TOD & Traffic conditions) Page 32 of 47

34 7.0 DATABASE REQUIREMENTS 7.1 Phase One Failure Management Model Global Parameters Table 7-1 Global Parameters Variable Default Description Thresh_High_Vol Dynamic Highest Possible Volume Thresh_High_Occ Dynamic Highest Possible Occupancy Thresh_High_V/O Dynamic Highest Possible Vol/Occ Thresh_Low_Vol Dynamic Lowest Possible Volume Thresh_Low_Occ Dynamic Lowest Possible Occupancy Thresh_Low_V/O Dynamic Lowest Possible Vol/Occ Suspect_to_Soft Dynamic Number of consecutive minutes a detector in the "suspect" status must fail the threshold test to be "Soft failed" Soft_to_Ok Dynamic Number of consecutive minutes a detector in the "soft failed" status must pass the threshold test to be reclassified "ok" Max_Fail_to_Hard 120 Number of times during the day a detector has failed its threshold test to be reclassified "Hard failed" Soft_to_Hard 30 Number of consecutive minutes a detector in the "soft fail" status must fail the threshold test to be reclassified "Hard failed" Percent_Good_Data 50 Percentage of good data necessary in each class to allow failure management to proceed The parameter Percent_Good_Data represents the amount of "good" data that must be present in each freeway section to perform dynamic failure management. The parameters Suspect_to_Soft, Soft_to_Ok, Max_Fail_to_Hard, and Soft_to_Hard represent the number of times the threshold test must fail and/or pass to change the state of a detector. The Suspect_to_Soft and Soft_to_Ok parameters are also dynamic based upon freeway conditions. A typical set of dynamic values for the volume threshold tests are shown below. Note that these tables are the same for all detector classes but the suspect_to_soft and soft_to_ok values only appear in the occupancy_threshold_table. Page 33 of 47