Inventing new ways to enhance the safety and reliability of autonomous vehicles

When we think about hitting the road in our cars, our first thought may not be that fellow drivers are particularly safe or careful—but human drivers are more reliable than might be expected. For every fatal car accident in the United States, motor vehicles travel over a hundred million miles on the road.

Human reliability also plays a role in how autonomous vehicles are integrated into traffic systems, particularly around safety considerations. Human drivers continue to surpass autonomous vehicles in their ability to make quick decisions and understand complex environments: autonomous vehicles have been known to struggle with common tasks, such as taking on-ramps or off-ramps, or turning left in front of oncoming traffic. Despite these enormous challenges, the future adoption of autonomous vehicles could bring huge benefits, such as clearing congested highways; increasing independence and mobility for non-drivers; and promoting driving efficiency, which is an important part of fighting climate change.

MIT engineer Cathy Wu envisions ways that autonomous vehicles could be deployed with their current shortcomings without experiencing a degradation in safety. “I started thinking more about the constraints. It’s very clear that the main barrier to the deployment of autonomous vehicles is safety and reliability,” says Wu.

One way forward may be to introduce hybrid systems, in which autonomous vehicles handle easier scenarios on their own, such as highway cruising, while shifting more complex maneuvers to remote human operators. Wu, who is a member of the Laboratory for Information and Decision Systems (LIDS), the Gilbert W. Winslow Assistant Professor of Civil and Environmental Engineering (CEE), and the MIT Institute for Data, Systems and Society (IDSS). , compares this approach to air traffic controllers on the ground directing commercial aircraft.

In a paper published on April 12 IEEE Transactions on RoboticsIn this paper, Wu and co-authors Cameron Hickert and Sirui Li (both graduate students at LIDS) present a framework for how remote human supervision can be scaled up to make hybrid systems efficient without compromising passenger safety. He noted that if autonomous vehicles were able to coordinate with each other on the road, they could reduce the number of moments humans needed to intervene.

Humans and Cars: Finding a Balance That’s Just Right

For the project, Wu, Hickert and Li sought to tackle a maneuver that autonomous vehicles often struggle to accomplish. They decided to focus on merging, especially when vehicles use the on-ramp to enter the highway. In real life, merging cars must speed up or slow down to avoid crashing into cars already on the road. In this scenario, if an autonomous vehicle was about to merge into traffic, a remote human observer could momentarily take control of the vehicle to ensure a safe merger.

To evaluate the efficiency of such a system, especially while guaranteeing security, the team specified the amount of time each human observer would be expected to spend on a single merge. They were interested in understanding whether a small number of remote human supervisors could successfully manage a large group of autonomous vehicles, and to what extent this human-to-car ratio could be improved while still being safe. Covering every merger formally.

With more autonomous vehicles in use, there may be a need for more remote supervisors. But in scenarios where autonomous vehicles are coordinated with each other, the team found that the cars could significantly reduce the number of steps humans need to take. For example, a coordinated autonomous vehicle already on a highway could adjust its speed to make room for a merge. The car completely averts a dangerous merging situation.

The team validated the ability to securely measure remote observations in two theorems. First, using a mathematical framework known as queuing theory, the researchers devised an expression to capture the probability of the number of observers failing to handle all the merges from multiple cars piling up simultaneously. In this way, the researchers were able to estimate how many remote observers would be needed to cover each potential merger conflict, based on the number of autonomous vehicles in use. To assist cars attempting to merge, the researchers derived a second theorem to measure the effect of cooperative autonomous vehicles on surrounding traffic in order to increase reliability.

When the team modeled a scenario in which 30% of the cars on the road were cooperative autonomous vehicles, they estimated that a ratio of one human observer for every 47 autonomous vehicles could cover 99.9999% of merging cases. But this level of coverage drops below 99%, an unacceptable threshold, in scenarios where autonomous vehicles do not cooperate with each other.

“If the vehicles were to coordinate and basically stop the need for supervision, that’s really the best way to improve reliability,” says Wu.

hovering towards the future

The team decided to focus on fusion not only because it is a challenge for autonomous vehicles, but also because it is a well-defined task associated with a less challenging scenario: highway driving. About half of the total miles traveled in the United States occur on interstates and other freeways. Since highways allow for higher speeds than city roads, Wu says, “If you can completely automate highway driving … you can give people back about a third of their driving time.” ”

If it became possible for autonomous vehicles to cruise unopposed for most highway driving, the challenge of safely navigating complex or unpredictable moments would remain. For example, “you (must) be able to handle the beginning and end of highway driving,” says Wu. You will need to be able to manage the time when passengers zone out or fall asleep, rendering them unable to quickly take control should the need arise. But if remote human observers can guide autonomous vehicles at critical moments, passengers may never need to touch the wheel. Apart from merging, other challenging situations on the highway include changing lanes and overtaking slow cars on the road.

Although remote supervision and coordinated autonomous vehicles are hypothetical for high-speed operations, and not currently in use, Wu hopes that thinking about these topics can spur development in the field.

“It gives us some more confidence that the autonomous driving experience can happen,” says Wu. “I think we need to get more creative about what we mean by ‘autonomous vehicles.’ We want to give people their time back safely. We want benefits, we want something strictly Do not want to run autonomously.”

This story is reprinted courtesy of MIT News (web.mit.edu/newsoffice/), a popular site that covers news about MIT research, innovation, and teaching.