What’s notable about this is the removal of the human from the car. For the outside public, it’s very hard to measure the level of progress of a robocar team. Everybody publishes very nice videos of their cars taking drives and solving various problems. The trouble is you can make a video like that at almost any level of progress, if you cherry pick what you show. As such, we have to measure teams by what risks they are willing to take and how many people they will allow to see all facets of operation.
The decision to go with no human in the vehicle means there was a big presentation by the team to the board where they showed the vehicle was good enough to release in this way, with members of the public and nobody to grab the wheel or hit emergency stop if a problem arises. That tells us that the team made a convincing case, and the quality is good — or perhaps that the team is reckless, which we will know soon enough. Baidu claims 32 million km of operations to date. Baidu states that while there is remote monitoring, they have roughly 2-3 vehicles per remote monitor, so it’s not a 1:1 ratio.
The vehicles must do pick-up/drop-off at designated stops, rather than just anywhere there’s a free curb the way human drivers do. “PuDo” is its own problem which not all teams have solved yet. (Cruise got in trouble for just doing PuDo in the street without pulling off, though at night this is common for taxis.)
The other measure of the team’s own self-appraisal of how far along is whether they will allow the public to see random rides. Again, it’s not so hard to take a guest member of the press along on a pre-planned and well tested route. If you allow members of the public to ride any where, any time, you show you are confident this will work. Some teams require the riders to sign NDAs and not to make videos. More confident teams have allowed anybody to release these videos. Again, this says the company’s own testing has told them their vehicle will not embarrass them in the videos. Baidu says riders may make and publish videos of their rides, so those will be interesting to see.
It is not enough to allow this of course. Tesla
To a minor extent, this also means they have convinced regulators of this, but the truth is the regulators are not really capable of assessing the quality of a robocar. Even the teams are figuring out exactly how to do that, but they are the only ones with much inkling. What they dare to do shows what their own evaluations have said.
In the USA, Waymo has been operating vehicles in Arizona with no supervising driver for several years now. More recently, Cruise began such operations at night in a limited area of downtown SF, and Waymo also began operations there at all hours, but has not started fully uncrewed service.
The ability to charge money is not a major step, though it has often touted highly. Nobody is trying to run these services as a business as yet. Charging money lets them see how the public reacts to the service when they have to pay for it, and to experiment with other types of charging. At present, most services are just doing charging similar to or slightly less than Uber