Driving Modules

Lane keeping

The basic thing in any road driving is lane keeping. Most self-driving cars rely on special infrastructure by the roadside to steer along the way. However, our research concentrates on making the vehicle learn to drive based only on the visual information picked up by the car camera.

Traffic signals

The next challenge in self-driving is the recognition of signals. The scene picked up by the front camera of the car is analyzed to pick up blobs of light. These are then classified as signals or non-signals; finally the color of the identified signals is extracted. We have obtained more than 95% accuracy in the identification of signals in the standard datasets.

Traffic signs

This offers a great challenge for most image recognition algorithms, given the complexity and variety of traffic signs. We have used several renowned algorithms and convolutional neural networks to achieve a very high accuracy in training the autonomous driving system to identify the traffic signs placed along the roadside.

Obstacle avoidance

True, obstacles are a part of life. There can be no true driving without avoiding obstacles. Our Convolutional Neural Networks have learnt the art of detecting obstacles on the road relying mostly on visual information. They are also capable of estimating the level of danger posed by a given obstacle, relying on fuzzy calculations. 

Gonken driving school

Before we run the RoboCar on the test driving course, we test the performance of the learning agent in the simulation environment. You can test your driving skills in this virtual environment. Along with our program, you will need the gamer's steering and pedal controls to enjoy the fun! 

Please reload

 2018 Gonsalves LAB. All Rights Reserved.

Copyright