Unfortunately VCs were much less interested in funding R&D of rolling autonomous robots in the past few years, so they aren't going to happen all of a sudden when the world actually needs them.
I do agree though that they make a lot more sense than drones, for several reasons: (a) power consumption / mileage per charge (b) carrying capacity (c) safety in populated areas (d) privacy when flying by buildings (e) lack of suitable landing areas in urban centers (f) noise in residential areas.
Source: I am the co-founder of one and aware of internals of several others
Safety I wonder. If they are only carrying 5lb packages having them fall from the sky sound bad but the frequency of rolling robots causing car accidents or running over people or getting smashed by locals might make it more safe.
Car accidents typically involve a car running into a robot, which might obliterate the robot but would seldom cause injuries to the driver. Delivery robots would almost exclusively operate on sidewalks of 25 mph, maybe 35mph roads. Running into someone is also unlikely to cause serious injury if the robot is kept to a reasonable size and weight limit. Obstacle avoidance and bump sensors are trivial to implement. Drones, on the other hand, can take out eyes and cause severe injury if they impact someone in the face with spinning blades, and crash avoidance mid-air involves an advanced control loop as opposed to just cutting power to motors.
Well, I agree. The idea of a UAV with a 2m wingspan landing somewhere near my house to deliver a 5kg package sounds like it could cause a lot of injuries. The delivery vehicle has to be much larger than the package and it has to land near a spot that is accessible to humans so they can pick the package up. As soon as you have flying drones they will have to get near other humans. Ground robots have a huge advantage in that they are less dangerous to humans. Do you really think something that is designed to constantly cooperate with humans is going to fare worse than something that will be near one for a minute? It's very likely that the UAV companies don't test their landing programs sufficiently.
I expect the delivery fleet of the future to look like an Amazon Locker on a self-driving e-bike trailer frame. You get a ping when your package is on your block, and use your phone to open your locker.
If costs get low enough, it might be possible to do multiple delivery attempts per day. Something like "your delivery device will be on your block at 7:00, 9:00, 12:00, 3:00, 6:00 and 8:00".
Or, more optimally, you would configure delivery windows and tell it when someone will be available and they schedule one or two delivery times in that window.
Another factor could be weather conditions. Heavy rain/snow/wind could all affect range or even accuracy (heavy precip can interfere with GPS to some degree). Drones also probably work better in a suburban environ, but less in a dense urban area or less dense rural area. Also, maps can be terribly slow to update. My home is in a Chicago suburb, about 8 years old now. Google maps still cannot accurately plot my address. Personally, I dont look forward to "we dropped off your package where we think your address is" and yet the package is a few hundred yards away. I'm also reminded of one of my favorite fuckups, this by USPS: my Mom misquote "street" instead of "drive" on my address, street number, name, city, state and zip otherwise being correct, and the overnight package landed in Florida instead of Illinois...
There's just no reason to waste the energy, the coordination effort or take on the risk.