Estimation of the Collision Risk of Autonomous Vehicle using Clustering

In this paper the clustering of different road-shapes like straight road or T-junction (i.e. three way road) or cross junction (i.e. four way road) has been presented using TFSOM×SOM algorithm. A trajectory path of a vehicle on different shapes of roads has been drawn which describes the degree of risk of collision on the road. The main aim of this work is to classify the different road-shapes and also to anticipate the possibility of the density of objects on the basis of different kinds of shapes of the road. The dependency of risk of collision is on the density of objects available on the road i.e. if number of available objects is more, the risk of collision is excessive and vice-versa. This prediction of high risk and low risk of collision can be used to set an alarm to alert driving system to avoid collision.