A new patent filing that has been uncovered describes the company’s efforts to deftly (and autonomously) navigate certain pickup and drop-off scenarios that could be handled with aplomb by human drivers, but could potentially stymie an AI driver. For example, a human would be able to quickly ascertain that an area under construction might not exactly be a good place to pick up passengers, or that stopping on the side of the road when traffic is whizzing by at 55 mph is not exactly a good idea.
Google’s patent gets around these predicaments by analyzing a requested pickup or drop-off location, which are then sent to a centralized dispatching system. If the user-defined locations are not in convenient or inherently safe locations, a list of suggested locations nearby would be provided. The dispatching system will also use a scoring algorithm to spit out the optimum pickup/drop-off locations based not only on the difficulty of the autonomous vehicle reaching the passenger, but also on what obstacles the passenger might have to travail to reach the vehicle:
For example, if a predetermined location on one side of a highway is closer in distance to a received location, but because of the difficulty involved in walking there from the received location, the scoring may identify another predetermined location that is farther away from the receiving location, but easier to reach by walking.
And problems can also arise at drop-off, in which an autonomous vehicle may be forced to provide an alternative location different from what was input by the passenger. The patent application describes:
In one example, the problem might be that the vehicle is no longer able to reach the destination location. This information may be provided to the vehicle’s computing device by way of traffic condition updates received or a traffic condition service.
The dispatching system could also be careful of avoiding road hazards such as potholes, slippery ground, and uneven or particularly high curbs that may cause problems for passenger by using onboard sensors.
This autonomous taxi service would be a perfect fit for Waymo’s fleet of 100 Chrysler Pacifica Hybrid self-driving minivans, which are about to start hitting the road. The endgame would be to replace human drivers completely, which of course takes a huge chunk out of the cost equation (although we wonder how Microsoft CEO Satya Nadella feels about that possibility).