This Indian Is Building World's 1st Self-Driving Car Safety System With Human Input

Autonomy vehicles are still a work in progress and cannot be completely relied upon just yet. A team of Indian American researchers aim to change this using human inputs.

MIT and Microsoft researchers have collaborated to create what will fill in the ¡°blind spots¡± in self-driving technology. For this, the concept employs human inputs, which will eventually help the autonomous technology to avoid any real-life errors.

The model spots anomalies in real-life situations which are not the same as the training examples feeded to the self-driving systems.

Representative Image (Reuters)

Also read: Your Amazon Order Might Soon Be Delivered By A Self-Driven Robot Named Scout

"The model helps autonomous systems better know what they don¡¯t know," says first author Ramya Ramakrishnan from Computer Science and Artificial Intelligence Laboratory at MIT.

The concept helps when the self-driving programs face a real-life situation which is different from what they have learnt in training simulations. Human inputs are hence used by the system to bridge the gap between the two.

The research was validated using video games, wherein human simulation corrected the learned path of an in-game character.

Representative Image: Reuters

Also read: Machine Kills Machine: A Speeding Autonomous Tesla Knocks Off A Humanoid Robot Standing On Its Way

Further, the concept is integrated into a conventional training program for the self-driving programs for vehicles and robots alike.

Ramakrishnan further added "When the system is deployed into the real world, it can use learned model to act more cautiously and intelligently."

(With inputs from IANS)