Road to Teslabot and Last Mile Delivery Robots

The last mile of a package delivery is the most difficult due to dynamic, uncertain environments both inside and outside of the vehicle. A delivery truck can start completely organized and packed with boxes. As packages are individually delivered, however, the neat stacks can slide or fall over as the vehicle moves. Outside of the vehicle, urban terrain varies based on both location and building type, e.g., delivering to a house with a porch versus an apartment complex that needs to be entered. The task is further complicated by the fact that unloading packages from a vehicle often specifies a different set of design criteria from dropping packages off at a desired location. As a result, combinations of robots with separate specialities are being explored, such as, manipulators on top of quadrupeds or manipulators cooperating with wheeled robots. These systems, though, come at a steep cost.

A Last mile robotic delivery project (LIMMS) has an analysis of robotic delivery. LIMMS criticizes legged robots systems. LIMMS says legged robots have historically had a higher cost of transport which reduces the number of deliveries performed before a required recharge. They tend to occupy large amounts of space that could be devoted to more packages.

LIMMS fills a niche between the general purpose and taskspecific robots. While it is suited mainly for manipulation tasks, it still remains adaptable for application on most other delivery tasks. The compact design allows multiple LIMMS to replace larger, complex systems.

Nextbigfuture believes fully functional humanoid Teslabots for delivery would be drop in replacements for human workers. Minor differences in wheeled robot efficiency would pale in significance to being able to be used in all human circumstances. Carrying loads over stairs and rough terrain so all houses and apartments can be handled is far more important.

Humanoid robots that can multiply the level of labor in factories would multiply factory production by hundreds of times. Humanoid robots for delivery with self driving cars would drive down the cost of supply chain logistics to 20% of current levels. This would put us onto the path of multiplying the equivalent of human labor by thousands or millions of times.

There would be replicating factories to send civilization into a massively exponential future.

Current practical solutions usually only solve a subset of the actual problem. For example, Starship robot is a six wheeled delivery robot that requires couriers or businesses to place packages in its cargo bay and recipients to remove the package on arrival. The robot is only responsible for package transportation. Other wheeled robots handle even more specialized tasks, such as Hikrobots
which operate in warehouses and sort packages by moving the shelves they are stored in.

Agility Robotics has its Digit bipedal robot. In 2019, they envisioned working with Ford on last mile delivery.

6 thoughts on “Road to Teslabot and Last Mile Delivery Robots”

  1. UAVs for small to mid size packages. Route could be optimized daily for loading, take off location and time, and return. Delivery truck may not even have to stop. Different size drones for differing weights.
    Apartments could have rooftop delivery, electronic door code access for larger buildings. Houses could have backyard delivery, which would be more secure.

  2. It’s a beautiful illustration of the early case for mass produced humanoid robots. Teslabots also wouldn’t need self driving delivery vehicles, like a human they could both drive an ordinary truck and carry packages to the door. They could also load the vehicle. They might of course start off doing only some of this – but this is a good example of a well suited early task.

    • Driving is a very complicated and dangerous operation so the operation of the self driving system must be perfect to many ‘9th’ (99.99999…%).
      If you look at the autonomous driving (AD) operation, you can divide it to 3 parts:
      1. Sensors: While humans can drive with just 2 eyes, they have many areas they can’t see, even with mirrors. Placing multiple cameras all around the car will give the AD much better chances.
      2. AI: The Humanoid Bot will probably has similar AI chip as the car so it can drive
      3. Controls: The interface between the AI and the car will be much better in an embedded AI then a bot using the driving wheel and paddles.

      Also the Bot will take significant space if used as a driver.

  3. Boston Dynamics first version of their handle robot (chicken legged wheeled feet robot with arms), similar to Digit, I felt was a decent compromise in terms of mobility and matching form factor to environment design. The LIMMS illustration also showed a wheel footed robot (not sure if you should call that a cooperative biped or a quadruped?)

    Also, Brian, it’s considered rude to direct link to the PDF on ArXiv, due to risk of updates to a paper. Please link to the abstract landing page.

    • Yeah, I’ve long thought it obvious that walking robots should have powered wheels for feet. The energy savings on any surface which would permit the wheels to be used would be phenomenal, and such surfaces are all around us, in any inhabited area.

      Sure, picking one or the other seems more purist, but combining them IS the practical approach.

Comments are closed.