Cruise and its autonomous driving technology

• Road to SIM, converting real scenes into editable virtual scenes. NPC (Non-player character) AI - comes from the concept of games, giving people and vehicles in the scene movement. World Gen, large-scale and rapid expansion of virtual worlds of different cities, and supports different light and time weather. Others also include virtualization of car dynamics and different sensors.

Of course, the construction of the autonomous driving metaverse does not mean that testing is not needed. Cruise CEO said in an interview that the principle of establishing the autonomous driving algorithm actually requires the definition input of user usage scenarios. So where does the user scenario definition come from? It comes from three aspects:

1. Road test and data collection

2. Engineers’ design

3. Virtualize extreme situations based on the road test data

Therefore, road testing is still very necessary. It affects the original requirement definition. A rich set of usage scenarios can ensure that the algorithm and subsequent work are correct. Development process tools

The development of autonomous driving, like vehicle development, actually requires a complete process and tool chain to ensure the efficiency of the many people involved in the development (to learn about the vehicle development process, click Vehicle Development Process - Final Chapter). Cruise claims that its tool chain covers everything from physical road tests to cloud code. The entire tool chain integrates data and automatically transfers data, allowing road test data to be transmitted and analyzed in real time, providing engineers with algorithm learning and virtual verification, and multiple algorithm updates can be deployed to road test vehicles within a week.

His toolchain includes:

Starfleet and Drive are used for data and scenarios on the road test side. Thousands of automatic production can be collected and entered into the Webviz platform in one day. Tools such as logging, inspector, and framework can automatically extract high-quality data. The Simulation system uses the Hydra batch scheduling platform in cooperation with Google for virtualization. Galileo should be the algorithm management system. Cartographer is the Cruise Maps network platform that allows users to create, visualize, edit, and test high-definition semantic maps. It is the center of all map creation and maintenance workflows. Finally, it is released to starfleet to deploy road test vehicles.

It sounds wonderful. In modern automobile development, especially in the era of big data, tools are productivity. I have seen various problems with tools and processes that greatly reduce development efficiency. Even if engineers work overtime day and night, it is just a waste of energy and reduces innovation. This is very common in domestic companies. Engineers work overtime, which is the biggest waste. They use tactical diligence to cover up strategic laziness. Cruise's future direction, of course, from the above, if you ask about Cruise's advantages, it should be backed by the big trees of General Motors in North America and Honda in Japan, which can be quickly implemented. After all, the sales of these two companies' cars are very high, so Cruise is not short of money and is not afraid of implementation problems.

The fastest-to-be-launched project is the Origin Robotaxi, which can be used in travel and short-distance logistics, relying on the engineering capabilities and supply chain of GM and Honda. This project is a bit like the Purpose Build Vehicle series of Hyundai Kia Group (KIA's 2030 strategy), and also a bit like the skateboard chassis that is very popular in China (what is a "skateboard chassis"? What is the business model?).

Therefore, Cruise's current technical reserves are scalable on a large scale and can build a tool chain. Therefore, the information currently circulating on the Internet by Cruise also reveals its hunger for talent. After all, it is not short of money, and investors also hope to see the scale of Origin in the future, and even feed back to GM and Honda's internal autonomous driving technology.

Of course, in addition to the implementation of autonomous driving, internal algorithms and tool chains, Cruise, like Tesla, has taken the path of developing its own chips. The chips include AI vehicle-side chips for sensors and central processing, as well as AI learning batch data processing cloud chips.

So to sum up, Cruise's ambition, driven by General Motors and Honda, should be huge, and its business model and algorithm tool chain are ready for large-scale deployment.

So do you think Cruise is a good choice? Please vote to find out.

References

GM Cruise Investor Report - Cruise

Autonomous Driving Moonshot Project with Quantum Leap from Hardware to Software & AI Focus - Deloitte

Autonomous Vehicles: Navigating the legal and regulatory issues of a driverless world - MCCA 

Automated Driving Safety Report-GM

The autonomous car - A consumer perspective - Capgemini

Autonomous driving testing and deployment issues - CRS

Three ways Cruise HD maps provide an advantage for our autonomous driving - ERIN, Senior Product Manager, Cruise

Cruise autonomous driving decision-making and planning technology analysis - Zhihu autonomous driving tractor

Cruise under the hood video - Cruise