Minus Zero has revealed an end-to-end autopilot system designed to support navigation in dense Indian urban traffic with on-coming vehicles and narrow roads with or without lane markings.
Tested on busy streets in Bangalore, the company says that its vision-based system can handle unique obstacles such as animals, push-karts, small two-wheelers (also known as Yulu), while understanding their unpredictable behaviors. To enable such features in unstructured traffic, Minus Zero has taken an AI-first approach by training end-to-end foundational models that can learn navigation in self-supervised manner from large scale raw data without human labels, instead of traditionally used rule-based systems.
The autopilot system more closely utilizes bespoke foundational models and self-supervised learning to navigate some complicated scenarios using only cameras, and without relying on HD Maps. Minus Zero noted that its focus on using an end-to-end models stems from the generalization capabilities they offer in new scenarios and new obstacles, which the company says makes them scalable and suited for unstructured and dynamic traffic environments like those in India and other emerging markets.
While demonstrating various scenarios through its latest autopilot technology, Minus Zero is presently working to broaden its functionality, adding overtaking, unprotected turns, roundabout navigation, and more. The company noted that the system ultimately comes under the ‘Hands Off & Eyes On’ category, where a safety driver must be onboard and attentive to the road in front of them at all times while it is active.
The system is currently in the development and validation stages, though Minus Zero expects production to commence within the next two years alongside its OEM partners.
