Building a self-driving car that people can trust
Over the past year, we’ve seen a rise in robotaxis and autonomous vehicle use. Companies such as Waymo, Cruise, and Baidu have all made strong headway as industry pioneers. In China specifically, 2020 headlines regularly featured major autonomous vehicle announcements, such as the public launch of Baidu Apollo robotaxi services in the cities of Beijing, Changsha, and Cangzhou.
But despite the increasing visibility of robotaxis and the public’s broader exposure to autonomous vehicle technology, many people remain hesitant about the safety of self-driving cars. Nearly three in four Americans say autonomous vehicle technology “is not ready for primetime,” according to a poll from Partners for Automated Vehicle Education (PAVE). Nearly half (48%) say they would never get in a taxi or ride-sharing vehicle that was self-driving.
Building trust between human passengers and self-driving cars is fundamental to the public’s use of autonomous vehicles. The current lack of trust may stem from high-profile controversies and accidents involving passengers and autonomous vehicles. Tesla’s FSD (Full Self-Driving) vehicles raised concerns because its technology was released without proper safety-focused communications while requiring drivers to constantly monitor the road. Uber’s self-driving vehicles drove controversy after one test drive ended with a fatality when the vehicle failed to recognize a pedestrian. The risks and mistrust of autonomous vehicles make safety the top priority for companies and their passengers.
As the leading autonomous driving player in China, Baidu has made significant progress building public trust in its autonomous vehicle technology. Part of its success stems from the continual development of industry-leading safety features, such as a state-of-the-art artificial intelligence (AI) system and 5G-enabled teleoperation. But just as important has been its deliberate focus on developing these leading safety features and communicating the ways these technologies and best practices—such as information security and safety assessments—ensure safe travel by minimizing hazards and accidents.
So what role do these technologies play in improving the safety of self-driving cars? How did Baidu communicate around these features to quell consumer concerns? And what can other companies learn from Baidu’s journey to better develop widespread public trust in self-driving vehicles?
A reliable AI system
The Baidu Apollo fleet of nearly 500 autonomous driving vehicles is known for its reliability and track record. The self-driving cars have driven more than 7 million kilometers (4.35 million miles) with zero accidents and have safely carried more than 210,000 passengers. In 2019, Baidu’s autonomous vehicles drove 108,300 miles in California, with only six disengagements and no accidents or injuries. Similarly, Baidu’s 52 autonomous vehicles traveled nearly 468,513 miles in Beijing in 2019 without incident.
Simulation tests and a well-devised verification mechanism are the key to building a safe autonomous vehicle. A 2018 report that demonstrates autonomous vehicle reliability calculates that it would take 8.8 billion miles—with a fleet of 100 vehicles being test-driven 24 hours a day, 365 days a year, at an average speed of 25 miles per hour.
Simulation makes large-scale testing of autonomous vehicles practically possible. Each iteration of the Baidu Apollo system is tested for millions of miles every day in a virtual environment that now accommodates over 10 million simulation scenarios. Baidu researchers also introduced a method to augment real-world pictures with a simulated traffic flow to create photorealistic simulation scenarios.
Before testing on public roads, the system undergoes a series of hierarchy verifications, starting from a fully virtual environment to mixed reality to closed areas. Each module of self-driving cars including models, software, sensors, and vehicles will be fully examined.
Driverless backups: 5G-enabled teleoperation
The Baidu Apollo integrated AI system allows its vehicles to drive independently without a safety driver inside the vehicle. To ensure public safety in the extreme road conditions, Baidu integrated 5G-enabled teleoperation into its vehicles.
The 5G Remote Driving Service, powered by 5G networks, smart transportation systems, and vehicle-to-everything technologies, provides immediate assistance from remote human operators during emergency situations. The operators aim to ensure the safety of passengers and pedestrians while its non-autonomous driving mode is in use.
All remote human operators have completed more than 1,000 hours of cloud-based driving training without any accidents and can ensure the safety of passengers and pedestrians when the non-autonomous driving mode is engaged. Since the autonomous vehicle drivers can handle most road conditions, human intervention is a rare occasion. Currently, Apollo’s self-driving cars in Changsha and Beijing have 5G-enabled teleoperation that can enable them to test on public roads without a safety driver.
Regardless of the efficacy of integrated AI systems, systems such as Baidu’s 5G Remote Driving Service guarantees that a person is always available to intervene.
To further the advancement of safety in mobility as a long-term vision toward the development of an autonomous driving ecosystem, Baidu launched an automotive cybersecurity lab, becoming the only Chinese company to do so.
The lab researches automotive cybersecurity technologies, trends, and solutions for in-vehicle systems, car-to-car communications, the controller area network, and sensors. It explores best security practices in data protection, in-vehicle infotainment, reference software, and hardware designs, and countermeasures to fake signals that misguide autonomous driving systems.
Knowing vulnerabilities and potential cyberattacks are industrywide concerns. In 2019, Baidu joined hands with automotive industry leaders and released a 157-page white paper that outlined how to build, test, and operate a safe automated vehicle. Through its CTF Challenge contest, Baidu encouraged developers and white hat hackers to design protection mechanisms for autonomous driving systems.
The lab, widespread information sharing, and strong industry collaboration are key components to maintaining public trust in autonomous vehicles. No system is completely impervious to cyberattacks, as was proven when researchers demonstrated the ability to remotely take control of a Jeep Cherokee (and technically most modern vehicles), controlling things like steering, brakes, and windshield wipers.
Baidu’s emphasis on industrywide security research and collaboration not only improves the security of these autonomous systems, it also demonstrates that public safety and security are its top priority when vulnerabilities are inevitably exposed.
Operation safety assessments
The final way Baidu is constantly assuring the public that its autonomous vehicle systems are working effectively is through regular, in-depth safety assessments. Baidu Apollo has a full-time “safety assurance team” which aims to prevent robotaxi accidents. The team is responsible for executing various aspects of operation area safety assessments, operation platform management, safety driver training, and safety assurance mechanisms. Team members evaluate the external environment of autonomous vehicles, including how to set up the road network and how to set up and expand the site.
Like any piece of equipment that comes with inherent safety risks, failures will happen, and safety systems can always be improved. By constantly reviewing and revising the safety measures it implements across its autonomous vehicle deployments, Baidu can catch issues before they result in an accident while providing ongoing reassurance to the public that it’s taking their safety seriously.
Building trust moving forward
Building trust between the public and autonomous vehicles is fundamental to their mass adoption. As an industry pioneer, Baidu has focused much of its innovation on developing and demonstrating technologies and best practices that can minimize risk. Through these initiatives, Baidu has developed a strong foundation of public trust in China that other companies can build on moving forward.
The key is that technology innovation is only half the battle. Reliable safety features like AI systems, teleoperation options, and strong information security are critical. But communicating how those innovations not only improve safety but compare to the everyday risks we accept with manual driving, is perhaps even more important. This is the model for shifting public sentiment and trust in self-driving vehicles.
This content was produced by Baidu. It was not written by MIT Technology Review’s editorial staff.
Geoffrey Hinton tells us why he’s now scared of the tech he helped build
“I have suddenly switched my views on whether these things are going to be more intelligent than us.”
ChatGPT is going to change education, not destroy it
The narrative around cheating students doesn’t tell the whole story. Meet the teachers who think generative AI could actually make learning better.
Meet the people who use Notion to plan their whole lives
The workplace tool’s appeal extends far beyond organizing work projects. Many users find it’s just as useful for managing their free time.
Learning to code isn’t enough
Historically, learn-to-code efforts have provided opportunities for the few, but new efforts are aiming to be inclusive.
Get the latest updates from
MIT Technology Review
Discover special offers, top stories, upcoming events, and more.