Traffic incidents happen everywhere in the city which is an all-time problem in smart city networks. Traffic incidents can be divided into two types, recurrent and random incidents. Recurrent incidents are those incidents has a regular pattern and happen in certain intervals and recurrent incidents are predictable by traditional transport engineering and data science. On the other hand, random incidents are the most difficult to deal with such as a traffic accident.
Nowadays, intelligent transport systems (ITS) are implemented in many cities such as Sydney in a reactive way. ITS makes decision based on the data of the on-road sensors. These systems are capable of relieving the congestions caused by the recurrent incidents but can barely response to the random traffic incidents. Therefore there is a need for building proactive/predictive systems in response to the random incidents.
Traffic simulation is an old-fashion way of modelling traffic conditions in smart cities which can predict future traffic conditions by simply simulating to the future. During the development of big data analytics and machine learning, traffic simulation also gets evolved by utilizing big data analysis to do self-calibration and to make better prediction of the future traffic conditions.
We are building a simulation framework in order to understand the impact of random incidents in a network in different transport modes. Historic data (such as ITS sensors data, population survey data, opal card data, motorway toll data, traffic signal data and so on) will be used to calibrate the simulation model and make the simulation model as close as the real world traffic. After calibrating the simulation model, it will be used for trials to predict the impact of the traffic incidents which can be presented as total travel time, total delay time, traffic flow, traffic density, and travel speed.
Based on the predicted duration and the predicted impact of the incident from other machine learning projects, several response plans will be tested in order to alleviate the congestion caused by the incident. Then the best response plan(s) will be suggested by this module to help the traffic planners to make decisions.
Currently, there is no systematic solution for random incident management and response, and the traffic managers make decisions based on their own experience. Therefore, it is a big challenge as well as a urgent need to build a framework for the smart city traffic incident response.
By using the latest technique including data science, it is highly possible to build an accurate and fast response framework.
Currently we have built several models in order to showcase the impact of random incidents on different transport modes in a smart city.
Example 1: Pedestrian train station modelling
In this example, we use the opal card data to estimate the number passengers tap on and off in one station, and then the trains follow the real timetables. Therefore, we can simulate the situation of platforms in a train station. Besides, as we can simulate the trains’ current load, we can infer how many passengers will be able to get on a train and how many passengers will have to wait for the next train.
Example 2: Train Incident Impact analysis
In this example, we simulate the impact of a random incident in a train line. The video shows the dynamics of the train and the time-distance diagram. The incident happened at 7:26 and lasted for 10 minutes, even a very small incident like this one was causing 3 trains stopping in the morning peak hours. You can image there would be more passengers waiting on the platforms in the downstream train stations.
Example 3: Sydney airport – incident impact using Aimsun micro simulation
In this example, we build the network surrounding Sydney airport and use the model simulated one artificial incident on the Robey Street from 7:30 AM to 7:55 AM in order to see the impact of this random incident. After the 25-minute blockage of all four lanes, we can see a grid lock for the surrounding road sections on all directions.