Waabi and Volvo Autonomous Solutions: developing autonomous trucks with generative AI

The future of AV 2.0 and its impact on autonomous driving

The evolution of autonomous transportation is marked by the advancement of AV 2.0, a system that represents a significant leap from its predecessor, AV 1.0. While AV 1.0 relied on smaller AI models and human intervention in the development process, AV 2.0 features a single AI system capable of performing all necessary driving tests. Rather than requiring large human teams, AV 2.0 uses models such as Large Language Models (LLM) to design autonomous systems with a high degree of precision.

The computational power behind AV 2.0 is provided by advanced technology such as Nvidia Drive Thor hardware, which powers Waabi Driver. This system is capable of managing the vast amount of data needed to train autonomous vehicles, although Urtasun warns that due to the “black-box” nature of AV 2.0, verifying that systems are functioning correctly is a challenge with considerable economic implications.

Waabi World: a virtual brain for autonomous vehicles

One of Waabi’s key innovations is the development of Waabi Driver, a “brain” for autonomous vehicles. This brain is trained through an advanced simulator called Waabi World, which generates digital twins of the world and performs real-time tests, allowing Waabi Driver to learn to drive autonomously without human intervention. With this high-fidelity simulation, the system can correct errors and improve its performance as it learns from environmental conditions.

An example of Waabi Driver’s effectiveness was demonstrated during a test where the system successfully drove in the rain, a challenge for autonomous vehicles that had not been specifically trained for this scenario.