In the context of holistic city-scale electro-mobility simulation large amounts of data is essential for validation and calibration. However, the number of sensors and their deployment is typically restricted. Furthermore, the size of the sensor network affects the computational complexity for calibration. This raises the question, how many sensors are needed and where to place them. In case an existing network of sensors is used, the question is which of these to use?
In order to monitor traffic, we need to place sensors such that the maximum of information on traffic flow can be retrieved with a minimum of sensors. For this purpose, it is important to identify key locations - in other words, the hot spots. Straight roads are less important since traffic flow does not change its direction. Intersections, on the other hand, are more relevant since traffic flows into different directions. Also, monitoring less frequented locations would not add much information.
We define a metric to quantify the importance of an intersection based on (1) the number of directions available in combination with (2) the degree to which the intersection is used by the commuter population. Simulation gives us a platform to test our method without real-world deployment. We use our SEMSim Platform to simulate the Singapore road network and derive the hot spots. The following video highlights the importance of potential sensor locations throughout a typical week day in Singapore.
As the business district in the south central part lights up in the morning as well as crucial intersections along the various passageways leading to it. In the afternoon, when people go home, a different set of locations become important. However, there are some locations which are important regardless the time of day. For example, in this particular scenario we can observe hot spots clustered around the intersection of PIE and Adam Road. Monitoring this area could provide information about the traffic flow from and to the central business district.
The quality of the results produced by this method depends on the quality of the input data. For the Singapore case study we used available data about the road network and origin/destination distributions derived from the Household Interview and Travel Survey 2012 which covers roughly 1% of the population. This amount of data may not be enough to come to any definite conclusions regarding hot spots. The work done evaluating the robustness of this method will be presented in a future blog article.