As robotics researchers we wish to take the challenge head on and aim to develop an AI model based on the spatial cognitive abilities of the Indian driver, which can indeed make vehicles demonstrate autonomous capabilities on highly unstructured Indian roads.
Our work entails several subsystems, such as Computer Vision,
Deep Learning, Localization, Path Planning, Control Systems, Sensor Fusion,
Middleware etc. with varying levels of complexity in each of them.
Self-driving cars (SDCs) have been pegged as one of the fastest-growing car markets in the world. As of now, a heavy amount of research is underway in this field, in controlled environments. With a disciplined road populace, uniform road quality, and structure, it is easier to come up with datasets and models needed for SDC deployment.
The discipline, or the lack thereof, of riders and drivers on the road, lack of strict enforcement of speed limits, rampant lane-jumping, signal-jumping, and helmet-less violations further adds to the problem with SDC deployment. Therefore, a different approach is required.
As robotics researchers and technology enthusiasts, the idea of an autonomous vehicle intrigues us immensely. More specifically, the kind of data sets and learning models necessary to train such a machine in a setting as complex and anarchic as Indian roads. We believe that progress in this direction will solve many fundamental challenges faced by SDCs on Indian roads and around the world.