This post was written by Ottopia’s CEO and founder: Amit Rosenzweig
At Ottopia, we are focused on making an automotive-grade teleoperation solution. in this article, we would like to explore why “advanced” teleoperation is the best way way forward.
We believe that the most promising solution for teleoperation should:
- Build on the most valuable lessons we’ve learned on the field; and
- Follow our ‘golden rule’: Let humans assist machines in decision-making, and let machines assist humans in execution.
The solution we are describing is something we like to call ‘advanced’ teleoperation. It is safer, more reliable and easier to scale than other options.
In the near future, picture yourself hailing a ride from a fleet of ‘robotaxis’. An AV picks you up at home, and your commute begins, safely and smoothly — until the AV drives up to a road construction site.
There are orange cones, workers in bright vests, old lane markings and newer, freshly painted lane markings. A police officer directs incoming cars, while cyclists and pedestrians travel to and from in multiple directions. This, for now, could be enough to challenge even the best self-driving AI out there.
In such an uncertain and complex situation, rather than risk provoking an accident, the AV should slow down and ‘call a friend’ for assistance.
But let’s pause for a second and ask ourselves this: What is the best possible way for a human to intervene?
A) Would the best way be for her to physically travel, as quickly as possible — say, in a motorcycle from an outpost nearby — to your location, jump in the driver’s seat, and take over the car for the next few blocks, until it can drive itself safely again?
B) Would the best way be for her to be remotely patched through and take complete control over the vehicle? That is one alternative, which, as we’ve highlighted before, comes with risks.
Or is there, C) an even better way?
Cue ‘advanced’ teleoperation. Let’s go back to that construction site, and to you, sitting on the passenger seat inside the AV.
As the vehicle slows down, a live feed of a human operator appears on the dashboard. On the other end, several miles away, the operator can see you, as well as the vehicle and its surroundings. She can hear what you hear. She realizes what’s tricky about driving down that street and past the construction. As you suspect, it’s easier for a human to make a smart choice about what to do in that situation.
Through her own set of controls, the human operator helps the AV make a decision about when to stop, when to advance, and which path to take. The vehicle executes these commands, with sensors and collision avoidance still fully engaged. If there is anything the human operator might have missed — e.g., a ditch hidden by the hood of the vehicle — then the vehicle detects and avoids it.
Slowly, surely, and safely, the AI and the human operator work together to navigate away from the construction site, and on to your destination.
What we like about a ‘smart’ system like this one is its adherence to our ‘golden rule’. It engages a highly skilled and trained human, only when she is needed, to resolve the situations that even the best AI out there can’t. But it also maintains the safety that comes from accurate sensing, navigation and actuation. These are all things that AVs are actually really good at doing, even when their decision-making in complex scenarios still isn’t perfect.
End-to-end execution is made as safe as possible, leaving minimal room for machine or human error.
What we like even more about this ‘advanced’ method of teleoperation is that it delivers what we at Ottopia are laser-focused on:
1. Safe navigation for passengers and pedestrians
2. Reliability and robustness to multiple external conditions
4. Scalability of the platform
If you’re wondering how exactly ‘advanced’ teleoperation delivers on all of the above… stay tuned.