Here’s the problem with the current batch of self-driving vehicles - they get into situations they can’t yet handle on their own and require a human to take over to get them out of it. It could be a construction zone, an accident scene, or any number of other unpredictable incidents. This is why most test vehicles are still required to have a human at wheel at all times, ready to take over. The vast majority of the time the car can handle itself just fine, and trying to stay vigilant just-in-case is exhausting (so much so that recently that Ford might have had a problem with their engineers falling asleep at the wheel!). This “human-at-wheel” issue is dramatically slowing progress on self-driving vehicles because they can’t go fully driverless (and realize the full benefits that this entails) until the vehicles can handle every situation.
Here’s my proposed solution: a cross between air traffic control and how the military flies drone missions thousands of miles overseas. Connect each self-driving vehicle and all its sensors and cameras to a central control hub where professional remote drivers will be stationed, ready to take over if needed. With new virtual reality tech, instead of a wall of screens in front of them, human drivers could use VR headsets for a much more realistic view of what’s happening with the vehicle.
For example, when a car gets into a construction zone that it’s unable to navigate, it alerts the control center that it needs help. A human driver then assumes control of the vehicle and remotely drives it out of whatever situation it was in. Once the vehicle is out of the problematic area, the remote human driver shifts control back to the vehicle. At some point, the hand-off could become so seamless that the humans in the vehicle wouldn’t even notice the transition.
With current tech, a self-driving vehicle can probably drive without help 95% of the time today, time a human is currently sitting at the wheel, bored and not doing anything. One remote driver could probably handle dozens of vehicles as long as there are enough drivers to handle unlikely situations where a higher percentage need drivers. This could greatly reduce the cost of riding with a self-driving car sharing service since you’re only paying for a fraction of a driver, and also rapidly increase public adoption. Trial runs could be done with empty vehicles in remote areas, or in testing facilities such as Michigan’s MCity, before any humans are put into potential danger.
There are issues with this solution, namely the lag in the connection between the vehicle and the control station. Even with current tech, I believe that this can be mitigated though, by slowing the vehicle down to reduce the effect of the lag and with software to help the human driver compensate. Also, as 5G networks roll out, the latency issue becomes less of a problem. Regardless, an in-vehicle human that hasn’t really been paying attention for that last two hours and is suddenly asked to take over probably doesn’t have a great response time either, at least initially. Having a remote driving control center solves a lot of problems and could help significantly in moving us into a self-driving future where vehicles never need human help.