Artificial intelligence is an incredible leap forward in technology, and its applications to transportation—especially autonomous vehicles—are nearly limitless. But as with other technological leaps forward, it’s impossible to see every possible edge case that might make AI more of a safety liability than a benefit. Enter teleoperation.

The solution of teleoperation shouldn’t be postponed until L4 or L5 vehicles are a reality. It solves needs now, enabling vehicles with Advanced Driver Assistance Systems to avoid obstacles. As an added benefit, teleoperation provides more jobs instead of taking them away.

Here are ten reasons why we need to involve teleoperations in autonomous vehicles in a responsible way, immediately.

1. Road construction. If you’ve ever been on a road trip in the summer, you know that the gestures of flaggers and signage in construction zones can vary widely. What’s more, construction sites can crop up quickly—and a vehicle’s mapping software can’t always stay updated with the latest info and may not know the latest construction sites. A teleoperator, however, is equipped with the necessary tools and knowledge to direct the vehicle safely through a road construction zone.
2. Inclement weather. AI can safely detect many hazards, but when unexpected hazards appear suddenly—like spring flooding washing out a bridge a few yards ahead—and an autonomous vehicle doesn’t know how to navigate around them, a teleoperator can turn the car safely around.
3. Agricultural and mining. The more specialized the vehicle, the more training it takes to operate. And although AI will eventually rise to that challenge, in the meantime, it’s easier and safer to have a human help operate such vehicles remotely than train AI to handle everything that could possibly come up.
4. Shipping presents similar challenges. Maneuvering a tractor-trailer into a tight spot at a loading dock of, for example, a decades-old warehouse, will be much easier to accomplish with a human teleoperator equipped with live video, satellite imagery, and detailed maps.
5. Crashes and crowds. Whether it’s a multiple car pileup or the dregs of a parade, an autonomous vehicle will do better with a teleoperator keeping an eye on people milling about, unexpected obstacles in the road, and other traffic.
6. Pedestrians. This doesn’t just mean a family out for a stroll with the dog. A human teleoperator can be invaluable in discerning among, for example, a police officer directing the autonomous vehicle to cross an intersection against a stoplight or a school crossing guard protecting children in a crosswalk.
7. Safety regulations. The federal government does not yet have nationwide safety regulations for autonomous vehicles; until it does, the onus to establish them falls to individual states. Teleoperations could help you comply with the most stringent state autonomous vehicle safety regulations.
8. Human fallibility. Although AI can detect and warn a human occupant of any number of dangerous situations, no amount of warnings will help if they go unheeded. A teleoperator can pay attention and intervene when a human occupant ignores warning signals or suffers a health event (such as a heart attack or stroke) that renders them unable to react.
9. Human-AI interaction. If existing AI interfaces such as Siri and Alexa have taught us anything, it’s that they don’t always understand what we say, even if we say it clearly. If a passenger has some sort of physical, cognitive, or speech impairment, it may be easier for a human teleoperator to interact with them than an onboard AI interface.
10. A skeptical public. Knowing that they have the option to interact with an actual human being can allay passengers’ concerns about purchasing or riding in an autonomous vehicle.

Food for thought:
100ms feedback delay is the limit of what humans can tolerate in closed loop control of a system (vehicle). Above that humans start noticing input delay and have issues controlling the system. Strangely, 100ms is also considered the fastest possible reaction time for human runners starting a race.