Abstract:
Large population, poor road infrastructure, and rapidly growing economies lead to severe tra c
congestion in many parts of the world. The problem is exacerbated by increased diversity in
vehicle types and poor adherence to lane discipline. Existing approaches for detecting tra c
congestion do not deal with the diversity of vehicle types and lack of lane discipline. We propose
a novel approach to detect congestion levels as per on the vehicle type. We present a thresholdbased
heuristic and an improved classi cation algorithm to infer the main regions of high tra c
in a city. We do transportation mode classi cation from the real time crowd-sourced smartphone
sensor data followed by congestion level detection. Our approach can be useful in congestion
management by suggesting optimal alternate route as per the vehicle type, thus saving travel
time and reducing fuel consumption and emission reduction. The proposed congestion detection
system was able to predict more than 80% major hot-spots of tra c classi ed as per di erent
vehicle types.
