Autobrains introduces agent-based AI architecture for ADAS and automated driving

Israeli autonomous driving company Autobrains has unveiled a new approach to vehicle intelligence, applying agent-based AI to both advanced driver assistance systems (ADAS) and automated driving (AD). The technology is already being deployed with global automotive partners in programmes aimed at mass-market vehicles.

Rather than relying on a single, unified model, Autobrains’ architecture distributes driving intelligence across multiple specialised agents, each designed to handle specific scenarios. Only the agents relevant to a given driving situation are activated in real time, enabling more efficient processing and reducing computational demands.

According to the company, this selective activation allows advanced driving capabilities to run on existing vehicle hardware and standard sensor configurations, avoiding the need for costly upgrades or platform redesigns by OEMs. The approach is positioned as an alternative to increasingly complex end-to-end systems, which have become more resource-intensive as autonomy capabilities have expanded.

Autobrains argues that its agent-based model offers greater flexibility and scalability, enabling continuous improvement of driving functions without relying on heavier hardware stacks. The company sees this as a shift in how autonomy is developed, focusing on structuring intelligence more efficiently rather than simply increasing computing power.