The Tesla Bot is happening.

During AI Day last August, Tesla unveiled the Tesla Bot. A humanoid robot that seemed like science fiction and after processing the announcement, many still imagined it to be many years away from being a reality.

This week, on Tesla’s Q4 2021 earnings call, Elon Musk made it very clear that creating a robot, is not an aspiration, but went as far as saying it is the most important product development happening at Tesla this year.

For many, myself included, this dramatically shifted the timeline of expectations. Rather than this being 5-10 years away, it suddenly seems that it could be here and be useful, in 2-3 years.

Having a near-term time horizon means there’s a long list of questions that arise that are important to explore, that should help us consider the potential the roadmap between here and a commercial product, the potential use cases, impacts on society, costs, and much more.

To explore these questions, this is the first post of a multi-part deep dive into the Tesla Bot, also known as Optimus (sub-prime).

What does the Tesla Bot prototype need to deliver in 2022 to impress?

When the Bot was announced, Musk said they’d have a prototype to show off next year. It’s now 2022 and while I wouldn’t expect it until late this year, when we see that prototype, it will provide our best opportunity yet, to accurately predict how far the commercial reality of shipping a Tesla Bot, that does useful things, really is.

Let’s imagine a dedicated event, ‘Tesla Bot Day’ happens in November this year. If we see a robot on stage, walking around in a human-like way, that’ll be a big start and a great box to check, but for it to be considered realistic in the next few years, we’ll need to see much more.

In the worst case, image the Bot was to fall over on stage, it would dramatically shift our projects to the negative, again, pushing the date back to a mid-term time horizon.

If however, the Bot not only walked, but actually had some speed about it, maybe a mild job, or slow run, it’d provide confidence that Tesla engineers have the basic mechanics of movement under control.

If they want to show off, they could demonstrate it climbing stairs, avoiding obstacles, and maybe even dealing with a human pushing it, that would be a seriously impressive start. This took Boston Dynamics years to master with their products like Atlas and Spot, but as we know, Tesla is such a different company, they iterate fast and I’d say that’s possible in just over 12 months from the announcement.

Boston Dynamics may be the closest competitor to what Tesla is trying to do. While the company is 30 years old, the Atlas humanoid robot, standing at 5-foot tall, was introduced in February 2016. In October 2018, they published a video titled ‘Parkour Atlas’ showing a single robot performing some basic jumps, but the most impressive to date was a video from August 2021, that showed two of their robots doing a full Parkour course, complete with backflips (not particularly useful, but impressive).

As it stands, Atlas uses 3 computers, and a pneumatic system, while Tesla will use electronic actuators to move the robot’s limbs. What we see from the Boston Dynamics robot Atlas, is still a large frame that has not gone through the miniaturisation of a commercial product.

As part of the ‘How does Atlas work‘ video, the company detail that they are still not at a stage where the robot navigates its world independently, rather what we see in the videos, is the robot executing a sequence of commands, quite different than what Tesla is aiming for.

This provides a good benchmark of where things are today in the robotics industry and for Tesla to expect, or us to expect of Tesla, that they will create a robot that surpasses more than 7 years of development, in effectively less than 18 months, is incredibly ambitious.

Now, if we see the Tesla Bot start to perform basic tasks, like picking up boxes from a shelf, identifying its contents using Computer Vision, then placing it on another shelf, based on the item type, that’s the sort of thing that would get the world, certainly, the commercial world would get really excited by.

For extra points, Tesla could create a larger scale task, like unpacking a series of boxes and having multiple Tesla Bots, work in harmony to move them from A to B. Taking this to the extreme, if they could demonstrate this being done at a speed equal to, or exceeding the speed at which humans could do the same task, then it’s time to get really excited about their ability to deliver.

Creating demonstrations for the stage is one thing, but it is quite another to show actual use cases, something like placing and removing a suitcase (or multiple suitcases) from the boot of a Model Y for example. I think the chances of us seeing complex tasks like this is extremely low this year, but would expect the level of importance that Musk placed on Optimus this week, may be possible in the next 2-3 years and this would certainly open our eyes to the dramatic utility afforded by a humanoid, compared to task-focused robots.

While Tesla Bot was announced to be 5’8 tall, it would not surprise me to see a prototype that was slightly larger, maybe 10-20%, to accommodate the fact now all the miniaturisation would be completed by this stage and allow for components to be slightly larger and better thermal envolopes. This would be iterated on over the coming years.

What’s probably also important, but not yet considered, is the amount of noise, or lack thereof, the Tesla Bot would create. We’ve seen a number of videos of robots being quite noisy in their operations. While this may not be a critical aspect on the factory floor, or outdoors, if the Bot is going to walk among us, it needs to do so at a certain volume level as to not annoy the general public.

During the announcement, Elon showed a slide that provided some detailed specs on the Tesla Bot internals, which could indicate they even had an early-stage prototype back then, to be confident they could achieve capabilities like strength and speed.

The Tesla Bot is designed to be 5 foot 8 inches in height, weighs 56kg (125lbs), can carry 20 kg (45 lbs), can deadlift 68kg (150lbs), extend its arms with 4.5kg (10lbs) of weight, and move at up to 8km/hr (5 mph).

It is not important that the prototype reaches or exceeds any of these figures this year, it is far more important that they can have a bot that looks like the design they showed and can walk around without falling down.

I fully expect Tesla to over-built these, then software-lock the Bot to make it safe and comfortable for humans to move around it. Could they make a robot that runs at 20km/hr, probably, but the usefulness of the robot doesn’t increase dramatically if it can run, instead you just increase the danger of being around it. The Bot does become much more useful if they focus on its ability to correctly identify objects and route an efficient path to where that object needs to go.

Something completely unknown is how the Tesla Bot would charge. Given its humanoid shape, it could potentially ship with a charger and plug itself in, that’d be handy. As we know wired charging is more efficient, but wireless charging is also an option.

If the feet of the TeslaBot contained a wireless charging receiver, similar to the bottom of your electric toothbrush, the bot could stand on a dock at the end of the day, charge overnight and be ready to go again in the morning, similar to charging your car, the cheapest charging could be done overnight in off-peak hours.

In the commercial world, power costs are negligible compared to the productivity gained by having a Tesla Bot, so it’s easy to see some companies wanting to run the Bot as close to 24hrs a day as possible. It’s interesting to think about a robot monitoring its own battery level, then returning to a charging bay, opening its chest, or legs, and replacing its own batteries, ready to go again for another 8 hours. I’d say it’s unlikely, but fun to think about. In reality, I think Tesla would rather sell you 2 or 3 bots to use on rotation while the other one recharges.

What about fast charging? Imaging Powerwall 3, many of which get installed in garages, could come with a charging cable that could be used by the bot.

Use the navigation below to see the next article in this Tesla Bot Deep Dive series.