Ahh but now you violate the cost assumptions of sequntially optimized delivery routes, exploding your fuel costs...a phenomenally bad idea. Tell me, are your delivery vehicles 10x more fuel efficient than a UPS van? Because that is the minimum range of efficiency that you'll have to acheive to make your plan viable, let alone cost effective. I'm not exaggerating either... you just added an entirely new dimension (delivery windows) to an already NP-hard Traveling Salesman problem. I wouldn't be surprised to hit a 100x increase in fuel costs for a pure scheduled delivery.
A UPS van sitting somewhere waiting is expensive because of the human inside. A driverless one costs nothing (except some refrigeration if applicable). Additionally residential have lower fuel costs for electric vehicles. It doesn't matter if the driverless vehicle takes 4 hours to deliver to an area when a UPS guy could have done it in an hour.
Note that UPS costs come from having the human. They need larger vehicles to absorb that cost, and then have to get the most bang for the buck in the time used. Driverless deliver can use different sized vehicles and time is far less important.
Except that your new idea (entirely distinct from your first idea where the autonomous driver could schedule its arrival to coincide with the consumer's whims) also is flawed. It is entirely common for an entire day's worth of fuel to be 1-3x the cost of the driver...especially so urban operations. Driving a minimal distance maximizes the driver's productivity, true...but it also minimizes the distance traveled, which minimizes the fuel consumed and the associated maintenance (which hovers around 40% of fuel costs). What you are now proposing isn't much better...instead of an autonomous vehicle running around catering to the scheduled whims of consumers, you have much more of them (which is a capital cost that scales linearly with peak usage during peak delivery times, while most sit idle for most of the day), running full round trips to a depot for every single delivery.
I'm all for creative solutions to problems. Your solutions just aren't creative enough. Operations Research scientists have been working on this exact problem for well over four decades...UPS probably employs 1/3 of the logistics-specialized OR professionals in the entire world. Saying these solutions are simple, or just require a little bit more thought, is not much different from saying that proving P=NP is simple and requires just a little bit more thought. In fact, it is exactly like that.
Why wouldn't the fuel be electricity which has no weight? And urban areas use considerably less of it (lower speeds etc). I didn't say the delivery vehicles do exactly one delivery at a time.
The point is that driverless vehicles can be different sizes than current vehicles, and are significantly less constrained because of no human driver. Consequently the delivery strategy can better match financials. They can do things like meet you as you get home. Or wait somewhere for 30 minutes to better match delivery windows. Or charge more for priority. Or deliver at 3am. Or bring vehicles from other areas to match changing demand.
Fewer constraints means more scope for optimisation.
