Autonomous Vehicle Development: Bosch & Cariad Deploy Test Fleets to Educate AI for Self-Driving Cars
In a groundbreaking collaboration, Bosch and Cariad, Volkswagen Group's software subsidiary, are jointly developing an AI-enhanced automated driving software stack aimed at Level 2 and Level 3 automation [1][2][3]. The goal is to make automated driving systems more accessible and safer for millions of motorists across mass-market to premium vehicles.
The software stack, designed for the Software-defined Vehicle (SDV) architecture of the Volkswagen Group, will enable drivers to take their hands off the wheel in various driving situations by performing perception, interpretation, decision-making, and action tasks with human-like AI-driven assistance [1][2].
The timeline for the project is clear. By mid-2026, Bosch and Cariad plan to provide the production-ready software stack for series project application [1][2][4]. First versions of the software are already being tested in fleets and trained daily using large amounts of driving data. Volkswagen Group will integrate the software into its new SDV architecture, starting with cars equipped from 2027 onward [1][2][4].
Bosch intends to offer this solution also to other global automakers outside the Volkswagen Group [1][2][4]. The stack is designed with end-to-end proprietary development, giving Bosch and Cariad full technical control over source code, security, data privacy, and safety standards, and allowing rapid innovation through source code optimization [1][2].
The AI stack analyzes urban scenarios and anticipates the behavior of other road users, using sensor modalities like cameras and radars. Sensor fusion of cameras and radar provides robust situational awareness for AI. In certain situations, AI prioritizes radar over cameras [1].
The architecture aims to keep decisions and actions of AI explainable, allowing results to be checked and documented for approvals. Level 2 of automated driving allows the system to steer, accelerate, and brake, requiring constant monitoring from the driver [1]. Measures against misinterpretations in AI, such as poor visibility or conflicting sensor data, are necessary for safety [1].
The development of the AI stack follows a cycle of drive, collect data, analyze, improve model, update software, and drive again [1]. In clear sight, AI relies more on cameras for depth perception. The alliance relies on a broad test and validation fleet in public traffic, with a focus on Europe, Japan, and the USA [1].
The AI stack developed by Bosch provides a basis for multimodal AI, including methods that couple language, images, and actions [1]. However, it's important to note that datasets for AI development must be correctly labeled [1].
In summary, Bosch and Cariad’s project aims to democratize automated driving by delivering a safer, smarter, AI-enhanced software stack that supports hands-off driving in multiple scenarios and will be widely available starting in 2026-2027 through Volkswagen’s new vehicle platforms and beyond [1][2][3][4]. Level 3 of automated driving allows the system to take over in approved scenarios, allowing the driver to briefly look away [1]. This collaboration is a significant step towards making automated driving a reality for the masses.
[1] Bosch Press Release: https://www.bosch.com/pressrelease/en/2021/bosch-and-cariad-are-developing-an-ai-powered-automated-driving-software-stack [2] Cariad Press Release: https://www.cariad.com/en/news/bosch-and-cariad-are-developing-an-ai-powered-automated-driving-software-stack [3] Volkswagen Group Press Release: https://www.volkswagenag.com/en/news/2021/01/bosch-and-cariad-are-developing-an-ai-powered-automated-driving-software-stack.html [4] Automotive World: https://www.automotiveworld.com/news-reviews/news-releases/bosch-and-cariad-developing-ai-powered-automated-driving-software-stack/
Technology plays a crucial role in the AI-enhanced automated driving software stack being developed by Bosch and Cariad. The software relies on sensors like cameras and radars for perception and situational awareness.
The goal of this collaboration is to democratize education in automation, making it more accessible and safer for millions of motorists through AI-powered solutions, which will be available starting from 2026-2027.
