Minus Zero develops vision-based autopilot for Indian roads

Bengaluru-based startup

has announced the development of an end-to-end autopilot system designed for

. The company claims the system has been tested on dense urban streets of Bengaluru, navigating without the need for lane markings and managing obstacles such as animals, push-carts, and small two-wheelers. The system operates using a vision-based architecture and does not rely on HD maps or LiDAR sensors.

Minus Zero uses foundational AI models trained in a self-supervised manner using large-scale raw data. Unlike rule-based systems that require manual input and precise environmental mapping, this approach allows the model to learn directly from unlabelled data and generalise to unseen roads and obstacles. The autopilot system fits under the ‘Hands OFF & Eyes ON’ category, with a safety driver required onboard during operation.

Focus on emerging markets and scalable deployment

The company sees its solution as suitable for emerging markets where traffic conditions are largely unstructured and road infrastructure varies. According to Minus Zero, while advanced systems like Tesla’s FSD or Mercedes’ Drive Pilot are operational in developed countries, the specific conditions in India present different challenges. Despite the majority of ADAS (Advanced Driver Assistance Systems) being deployed in developed regions, a significant portion of global road accidents occur in emerging countries, highlighting the need for context-specific solutions.

Minus Zero said it is already collaborating with a couple of large automotive manufacturers in India and is in talks with others to integrate this technology. The company targets production readiness within two years in partnership with OEMs.

Technology highlights and approach

Shift in approach: The system is based on foundational AI models, moving away from supervised learning and rule-based systems.

Hardware efficiency: The platform uses a vision-first approach, minimising dependency on expensive sensor suites.

Mapless navigation: The system does not rely on HD maps, which are currently not widely available in India.

Scalability: The AI models are trained to improve continuously, allowing the solution to adapt across different geographies and vehicle platforms.

Full-stack development:

Minus Zero has developed its own generative AI tools, simulation environments, and onboard software for deployment.

“After LLMs and AI agents for the digital world, next-gen foundational models for the physical world like autonomous driving, robotics, etc. are gaining momentum globally,” the company said in a media release. “With India taking a strong stance in creating a foothold in the AI race, this area is an opportunity where we can build in India for the world.”