Japanese automaker Nissan Motor has started testing its new driver-assistance system, powered by technology from British start-up Wayve, ahead of a planned launch in the Asian country during the 2027 financial year. The automaker recently demonstrated the system in Tokyo, using Ariya electric vehicles equipped with advanced collision avoidance features designed to assist drivers in urban areas.
Wayve, which received funding from SoftBank Group and Nvidia, opened a testing and development centre in Japan a few months back. Nissan, which launched its ProPilot system in 2016 and rolled out a second-generation version in 2019 to assist with highway driving, has not yet announced which models will come equipped with the next-generation driver-assistance system. However, Wayve is playing a crucial role in developing the Japanese automaker’s upcoming driver-assistance system.
Meet The New Player
The British start-up has quickly become one of the best-funded companies in the UK, while also gaining recognition as one of the few artificial intelligence pioneers based in the country. According to Alex Kendall, the founder of Wayve, who spoke with The Guardian, in addition to Nissan, the company is working with large manufacturers in Europe, North America, and Japan.
In fact, chip giant Nvidia recently signed a letter of intent for a possible USD 500 million investment in Wayve’s next funding round. The start-up has already raised USD 1.3 billion from investors, including Japan’s SoftBank, to fund its expansion in the United States, Germany, Israel, and Japan, as well as in London.
Nvidia is already providing one or two chips in each car using Wayve technology, and many more in the data centres used to train Wayve’s foundation model on vast amounts of driving data, including videos of drivers encountering real-world models.
Kendall founded Wayve in 2017 after studying deep learning for computer vision and robotics at the University of Cambridge. He told The Guardian, “We want to build a trillion-dollar company,” while adding that the company had reached “a real inflection point in the capabilities of this technology” that allowed it to learn rapidly how to navigate Tokyo’s crowded streets.
Wayve uses AI models that learn how to drive by ingesting huge amounts of video and driving data, while spotting patterns it can then replicate. This approach contrasts with some other driverless technology companies, which tried to program explicit rules into their systems. Nissan was already developing its own autonomous driving technology, including in London trials, but the opportunity to use Wayve’s expertise led to the breakthrough the British start-up needed.
Robotaxis controlled by rival software are already operating in some American and Chinese cities, although companies have faced problems with cars occasionally being unable to navigate unusual obstacles.
Wayve cars use a combination of cameras and radar, similar to the Nissan cars, which also employ a more expensive lidar laser sensor. Kendall said this would provide “a level of redundancy” in an affordable manner.
A Unique Technology
The company, which has a distinguished list of investors such as Microsoft, SoftBank, Uber, Virgin Group, Nvidia, and Eclipse, has based its driver-assistance system on the “AV2.0 Approach,” which is pioneering the start-up’s end-to-end AI Driving Model. This replaces the modular “sense-plan-act” architecture of the traditional AV1.0 approach with a single neural network trained on diverse data to convert raw sensor inputs into safe driving outputs.
The technology learns driving skills from raw, unlabelled data using self-supervised learning, thereby eliminating the need to curate expensive and time-consuming labelled datasets. Additionally, AV2.0 doesn’t rely on HD maps, which enables seamless expansion to new geographies through data-driven adaptations. AV2.0 allows Wayve to think differently about sensors.
This data-first approach is flexible in terms of sensor selection, giving OEMs (original equipment manufacturers) the freedom to choose hardware based on their needs. AV2.0 can adapt to operate on any type of vehicle, from passenger cars to delivery vans. Advances made in either vehicle type directly benefit the other.
AV2.0 has introduced a rapid, continuous, and seamless fleet-learning loop: recording data, training models, evaluating performance, and deploying updated models. The “Fleet Learning Loop” efficiently gathers real-world driving data from diverse fleets, processes it in a cloud-based training infrastructure, and converts it into refined driving capabilities.
The mechanism has been optimally designed to support the transition from “eyes-on” driving functions to “eyes-off” as driving data exposure builds verifiably robust automated driving capabilities.
Following the philosophy of “Responsible Model Development,” AV2.0 implements MLOps workflows for responsible model development, utilising innovative tools, processes, and pipelines to build, train, and deploy foundation models. Before “On-Road Trials,” the start-up rigorously tests its AI driving models across a vast array of simulated driving scenarios for rapid and comprehensive evaluation.
Under the “Safety 2.0 Framework,” Wayve has introduced a revolutionary approach that addresses safety through the lens of deep world understanding. Unlike traditional methods, “Safety 2.0” acknowledges that true safety comes from an AI that, like a human driver, interprets the driving environment naturally. The result has been the creation of an AI system that inherently comprehends the intricacies of the world and drives behaviours for safe navigation.
Utilising Generative AI, Wayve has developed a high-fidelity, predictive world model that understands the implications of the vehicle’s actions and ensures safe responses to other road users. The model’s understanding can be used to prevent unsafe actions. Wayve’s end-to-end (e2e) model architecture and active learning processes have been optimised to produce effective, fluent, and safe driving behaviours from unlabelled driving videos, showcasing strong generalisation across various geographies, vehicles, and sensors. The entire model can be further enhanced with multimodal data sources that provide additional information.
New methods have been created for dataset introspection and control, while harnessing the emergent concepts generated by the AI and transforming them into a framework for deeper analysis. Safety 2.0 also carries comprehensive datasets for diverse driving scenarios and filters out low-quality and risky scenarios.
The model also has its own introspection method, where it uses natural language and other information to evaluate the AI’s decision-making process. OEMs also get control and visibility over their data assets and learning processes to ensure that Wayve’s AI models achieve the highest safety standards.
Advanced AI Making Driving Safe
Wayve’s “AI Driver” is transforming driving automation for foreign carmakers. The start-up’s mapless, hardware-agnostic solution enables efficient software upgrades, unlocking advanced levels of automation from L2+ to L4 as Wayve’s core AI model evolves. The approach resonates with automakers’ goals of ensuring innovation, safety, adaptability, and scalability when fielding driver-assistance systems.
The platform is designed for universal compatibility, allowing straightforward integration into any vehicle. This empowers OEMs to easily adopt automated features across their entire vehicle portfolio, promoting a cohesive driving experience across the lineup.
The unique mapless AI solution also facilitates effortless adaptation to any environment, eliminating the need for detailed high-definition maps. This simplifies global deployment and lowers costs in the process, allowing for scalable expansion of automated driving technology. Wayve’s “AI Driver” has also been future-proofed by enabling seamless over-the-air updates to support different levels of driving automation.
The “AI Driver” offers a comprehensive suite that includes foundational driving models, safety mechanisms, APIs, data collection, and cloud-based monitoring and configuration infrastructure. This platform is engineered to optimise performance, configuration, and safety, meeting the high standards of the automotive industry, including hands-free driver assistance day or night, from city to highway, with minimal hardware requirements. The entire package of comprehensive warning and active safety systems has been made global NCAP (New Car Assessment Programme)-compliant.
To scale globally, Wayve has set very tough qualification parameters for its AI Driver: it must perform in places it’s never seen before, with no region-specific retraining and no HD maps. The start-up has launched the “AI-500 Roadshow,” a bold plan to take its single foundation model to 500 cities by the end of 2025.
Image Credits: Wayve