VIEWS: 6 PAGES: 4 POSTED ON: 5/20/2011
Microsimulation Modelling Report Index QLD-5 Location / route / area Gympie Road: Stafford Road to Edinburgh Castle Road Project description Aimsun Analysis Purpose of modelling To see the effect of upgrading an unsignalised intersection to a signalised intersection. The route choice model was used to determine the proportion of vehicles that change routes when the intersection is upgraded. Model developed by Sandra Lennie Microsimulation software used Aimsun NG Author of report Sandra Lennie Date of report 16/03/06 Model scope description Geographical extent Gympie Road: Stafford Road to Edinburgh Castle Road Years modelled 2006 Time periods modelled 6:45-7:45 and 4:30-5:30 Time periodic variations (profiles) in: No variation in flow demand Traffic demand Links Junction control Number of zones n/a Number of links 40 Number of nodes 9 Number of junctions 9 Number of traffic signals: 5 (all fixed time) Fixed time Vehicle-actuated Area traffic control system Networks Base Network Built up over scaled aerial photos Basic geometry As per existing road layout Intersection layouts As per aerial photos and intersection layouts Traffic signal controls As per STREAMS Other network features, e.g. 1. Roundabout Signposting 2. Giveway and stop signage at unsignalised intsections Ramp metering 3. Bus stops Adjacent lane interaction Lane restrictions Time dependent features Carparks Future networks Options were developed by signalising one intersection. A Sidra analysis was used to determine suitable timings. Basic geometry n/a Intersection layouts n/a Traffic signal controls Sidra analysis used (as above) Other variations from base network Vehicle and driver data Data type Default vehicle data used Queensland vehicles as supplied by David Stewart Additional or non-standard vehicles Vehicles used include cars, heavy vehicles, semi-trailers and buses used? Vehicle proportions Cars 95%, heavy vehicles 4% and semi-trailers 1% Headway Reaction time 0.75 Driver behavioural parameters, e.g. Reaction time at stop 1.35 sec Familiarity Aggression Awareness Base travel demand Source of raw data Manual traffic counts supplemented by STREAMS data Automatic vehicle counts No Manual vehicle counts No Classified counts No Manual turning counts Yes Counts from signal control systems Yes Counts from freeway management No systems Number plate survey Partial Roadside interviews No Mail-back questionnaire No Home interview No Commercial vehicle survey No Other sources No Base trip table estimation Method Traffic turning counts were input and then an API was used to extract an approximate matrix. Further finessing was undertaken Counts only Base travel demand data from above was used Synthesised from counts: Yes Observed Modelled Other Details of time dependent demand Peak hours were used in the model profiles used Future trip table estimation Method No change was required Growth factors Modelled Other Adequately defined in the brief? Work complies with the brief? Work adequately documented? Assignment details Algorithm c-Logit Route choice model was used, because it reduces the amount of “flip flopping” between time periods Cost coefficients Capacity was lowered on some routes and extra delay was forced onto back streets. This encouraged vehicles to use the main Arterial road (Gympie Road) Incidents No Signposting No Strategic Routes No Calibration Calibrated To The saturation flow rate at signalised intersections was calibrated and the turning counts were also calibrated Trip length distribution Observed volumes Calibrated according to GEH statistic. 85% of turning counts must have a GEH of less than 5 and the total number of vehicles to enter the network must have a GEH of less than 4. Maximum flows Saturation flow rate was calculated and compared with an average for the region. No Saturation surveys were undertaken. Queue lengths Travel times Other (specify) Validation Has the calibrated model been No validated against data not used for calibration? Validated against Observed volumes Maximum flows Queue lengths Travel times Other (specify) Model application Please discuss below in one to two pages the following issues: results of investigating different scenarios sensitivity tests undertaken extent of the variation from default parameters difficulties encountered and ways to overcome modelling issues comments on the general robustness of model outputs A stakeholder of the Kedron State School was concerned about the amount of traffic travelling outside the local primary school on Leckie Rd. They suggested that this traffic be encouraged to divert away from Leckie Rd to other parallel roads. It was thought that signalising Gympie/Edinburgh Castle Rd would reduce the overall trip delay for motorists travelling from Edinburgh Castle Roundabout to Gympie Rd (SB), because vehicles would be able to enter Gympie Rd more easily. A micro simulation model was created in Aimsun NG to determine the effect of upgrading Gympie/Edinburgh Castle Rd. The route choice model was used to identify the change in driver route choice due to the upgrade. The simulation model showed that upgrading Gympie/Edinburgh Castle Rd produced the opposite result from what was expected. When Gympie/Edinburgh Castle Rd was signalised, more traffic was encouraged to travel past the school along Leckie Road, because of extra delay at Gympie/Edinburgh Castle Rd. Traffic entering and exiting from Edinburgh Castle Rd was now forced to wait for a green signal, where previously they would have been able to select their own gaps in the Gympie Rd traffic. In turn this extra delay experienced at Edinburgh Castle Rd is detected by the route choice model and as a result other routes (such as Sadlier Street) become more attractive (because they now have less delay). In fact, after the Gympie/Edinburgh Castle Rd intersection was signalised, there was a 19% shift in the AM and 13% shift in the PM of traffic away from this intersection towards Leckie St and Sadlier St. Public transport along Gympie Rd was not significantly affected. Extent of variation from default parameters Variation from the default parameters was only in the route choice model. The „Capacity Weight‟ factor which penalises low volume routes was significantly experimented with to achieve the correct balance between volumes on the back streets and the main arterial road. The „Scale‟ factor which determines the proportion of vehicles that use the shortest route was also experimented with. The „Beta‟ and „Gamma‟ factors were changed to reduce the amount of flip flopping between route choice time periods. Individual link volume and delay parameters were changed to attract or deter vehicles from using individual links The basic method was to calibrate the global parameters first and then finesse the route choice model was making small changes to the individual link parameters. Difficulties encountered Significant difficulties were encountered with the route choice model and the public transport 1. Route choice model and the flip flop effect: The route choice model directs vehicles along the path with the least delay in each time period. This is best illustrated with an example. In the example, there are two alternative routes. The route choice model must direct vehicles to the destination using one of the alternative routes and the route choice model directs the majority of vehicles to the route with the least amount of travel time. In time period 1, the route choice model calculates that route 1 has the least amount of travel time and therefore in time period 2 vehicles are attracted to route 1. However, now route 1 experiences a lot of delay and in time period 3, vehicles are directed to route 2. This process repeats and the volumes of vehicles on route 1 and route 2 flip flops at each change in time period. This problem was moderated stabilised by calibrating the Beta and Gamma factors. 2. Public Transport: In the process of modelling this project a number of bugs were found with the public transport section of the Aimsun NG model. Particularly buses were finding it difficult to enter bus stops. These were reported to TSS and have now been fixed. Robustness of the model Robustness of the model is unknown, so the modelling report given to the client highlighted the changes between the different options rather than the absolute results.
Pages to are hidden for
"Microsimulation Modelling Report"Please download to view full document