The technological society Waymo, which is part of the Google universe, continues its tests on autonomous driving. In 2021 there was a major qualitative leap, with the expansion of operations in the city of San Francisco in California; this is where work is being concentrated on refining the system that allows cars to drive without a driver.
In the decade that American company hit the streets of San Francisco, five generations of the Waymo Driver self-driving system were designed. In particular it was the vehicle Jaguar I-PACE to provide the ideal platform for testing. Along the way, Waymo became the first company to operate a fully self-contained commercial freight service: Waymo One. The tests during this period have been assiduous, with a specific evaluation method in order to find the right formula capable of giving the necessary safety and confidence to motorists.
We read on Waymo’s blog: “After conducting our simulation and closed-circuit tests, we independently traveled over 2.7 million miles around the city to advance and validate our performance. These kilometers are essential for training the system, ensuring simulation realism, evaluating driver performance and continuously improving safety. We evaluated the software performance of the Waymo Driver in increasingly demanding driving situations, day and night, in different weather conditions, from the dense traffic of Masonic Avenue to the numerous pedestrians of Valencia Street and the road closures on Saturday nights. Finally, we started offering rides to community members – hundreds of people travel with Waymo One in San Francisco now, providing valuable feedback to help us refine and advance our service. As we work to remove drivers overseeing self-driving operations from San Francisco vehicles, we continue to evaluate three core elements: hardware, software, and operations.“.
As for thehardware, Waymo for each car in the fleet must validate the integration of the Waymo Driver sensors against the basic ones on the Jaguar I-PACE. Sensors need to properly evaluate the vehicle’s steering and speed control systems, particularly adapting to San Francisco’s narrow streets, curves and hills. They must promptly activate the computing platform that allows for a quick reaction in dangerous but typical situations in the city, i.e. when an emergency vehicle approaches or when someone passes a red light at a busy urban intersection. Sensors must act regardless of weather conditions, including phenomena such as fog, rainstorms or sandstorms. Finally, they must be robust enough.
From the side software, the complexity is enormous. The system must be able to read and better still anticipate scenarios full of information, such as complex intersections, narrow streets, constantly changing layouts and social interactions between drivers, cyclists and pedestrians. A commonly proposed approach to help autonomous vehicles avoid conflicts on the road is to insert strict rules into the driving software, such as “always stay at least 40cm away from any other object”. But this doesn’t work in dense cities, where vehicles have to regularly handle curves and tight roads. Waymo Driver has been designed to better adapt to such situations, where a real driver could invent evasive and effective maneuvers to shake off traffic. Another important issue is fog: the software must know how to decide how much to reduce the speed based on the density of the phenomenon.
Finally, the autonomous driving system must know how to be useful and quickly adapt to the operations that are required. In San Francisco, the Waymo One service can be updated with new routes or features, so it can adapt to more stops or unexpected situations such as sudden road works. At the base of everything there is a specific program that evaluates risk management, with the always open possibility to proactively learn to re-evaluate choices, like an algorithm that is constantly improving. This is the only way to avoid that the car is reduced to simply shuttling from point A to destination B, remaining at the mercy of adverse events.
#Waymo #Autonomous #Driving #Works #FormulaPassionit