Mobile Crowed Sensing based Intelligent Transport Systems

Mobile Crowed Sensing based Intelligent Transport Systems

In Ethiopia, there is a dramatic increasing of vehicle number especially in the capital, Addis Ababa. However, road Infrastructure growth is lagging behind growth in vehicle volumes. Better traffic management and more information to travelers can alleviate traffic woes.  

Every day, we are experiencing horrible traffic accidents as shown on Figure 1(a) and daily crowed and congestion Figure 1(b).

Current traffic monitoring systems use fixed-position traffic sensors (using either roadside cameras or magnetic loops embedded in the ground) to monitor traffic volume. However, the high cost of the equipment and its installation makes such a system available only on major roads.

Now that, many people carry mobile phones with a GPS receiver, mobile-phone-based traffic monitoring systems are gaining attention due to their low cost and large coverage (e.g., the Mobile Millennium Project by UC Berkeley).

Mobile Crowed sensing is new paradigm of mobile computing where anytime, anywhere computing as mobile phones are the ideal devices since they always with us, Internet enabled, locatable (GPS or other systems). New applications benefits from real-time location and place information. In the near-future, we will benefits from other types of sensing/ context information. A typical smart phone currently have ambient light sensor, proximity sensor, dual cameras, GPS, accelerometer, dual microphones, compass, Gyroscope. 

We proposed Mobile crowed sensing based Intelligent Transport System. We are mainly focused on finding sensing techniques and algorithms, congestion detection/alleviation and reporting on Ethiopian roads.

Traffic congestion predication - time serious forecasting -  a series of measurement from selected roads in Addis  Ababa. The goal is to make short-term predication of future values based on historical ones. Traffic reconstruction and predication based on real-time information from individual drivers. Input data consists of a stream of notification from 1% of vehicles about their current GPS locations in the city road network, sent every 10 seconds. Our algorithms receives this stream and predicts traffic congestion on the selected road segments from the next 30 minutes. Large volumes of data are involved in this task, requiring the use of scaleable machine learning methods.