Artificial Intelligence in Autonomous Vehicle for Collision Avoidance at Intersection

On-going Project: #1

Recent studies have reported that more than 2.5M (almost 40% of all the total) crashes in the United States are due to intersection collisions. To reduce the number of accidents, new developed systems must be able to exchange their dynamic information such as speed, acceleration, direction, relative position etc. with the other vehicles through Intersection Traffic Controller (ITC) in real time. For that, analysing or monitoring traffic behaviour at an intersection, such as collecting the traffic speed data, motion trajectory and counts for different kinds of traffic objects is very important. ITC will keep broadcasting a message all the time and each vehicle will communicate with the ITC in real time to let the ITC know about its current location. This ITC will calculate the shortest path of communication between the vehicles to provide faster and efficient communication between different networks and provide smooth traffic flow at intersections.

One possible solution is to use the ad-hoc networks for wireless communications at the intersection. Ad-hoc is the group of wireless nodes that communicate amongst themselves to form a network. But this leads to the problem of interference because when a node is transmitting to one of its neighbours, other neighbours need to be silent due to the interference in the medium.

The second solution is to use flooding as the simplest way to broadcast but that is relatively well for a limited small number of nodes, however, the performance drops quickly as the nodes increase.

The efficient way is to develop a centralised broadcasting system algorithm that will utilize the neighbourhood information to reduce redundant messages in a automobile Ad-Hok network. There will be Control Channel zone and Service Channel Zone in the model. The communication between the vehicle’s on-board unit and centralise broadcasting system will be started as soon as the vehicle enters in these zones. This system will provide the information from the intersection point to warn the driver about the status of the vehicles and the signal to avoid the collision. The node’s relative speed is an important metric in comparing different mobility models. The nodes in bidirectional mobility models have less correlation among each other, except for neighbouring nodes travelling in the same direction.

My teammates and I are using Python, MATLAB and Simulink for vehicle modelling and their communication. We are going to use Google’s gRPC as a communication protocol.

PS: If you have any suggestions or ideas related to this work then please comment below or dm me.