This Company Will Be The First To Crack Robotaxis
No one else comes close.
This Company Will Be The First To Crack Robotaxis
No one else comes close.
When I say robotaxis, you most likely only think of one name: Tesla. Musk’s revolutionary company has been publicly gunning to be the first to offer this futuristic self-driving technology since 2014. But, since then, Tesla’s repeated promises of fully autonomous vehicles have fallen utterly flat, and some questionable business practices and decisions from Musk have rendered their self-driving development program deeply controversial. But, another EV pioneer has been working on autonomous vehicles for almost as long as Tesla, just without the PR stunts: Rimac. The EV hypercar company has been working on autonomous driving since 2017 and received a €200 million fund from the EU to develop robotaxis. Well, they recently announced to the world their robotaxi company and vehicle, called Verne. When you look closely at the car and how Verne plans to deploy it, it becomes crystal clear that they are by far the best contender in the robotaxi race and could demolish Tesla and everyone else in the autonomous vehicle world. Let me explain why.
The car itself looks bizarre and amazing. It is the size of a typical hatchback and only has two seats, but these seats are at the back of the car. This allows it to have enormous legroom, massive reclining seats, a giant screen and a sizeable boot. The suspension is roughed to survive the harsh life of a taxi, but the body is isolated from the drivetrain, meaning shocks and road noise barely make it into the cabin. The car is classed as a Level 4 autonomous vehicle and, as such, has no steering wheel or pedals, giving its cabin the look of a luxury airline. It only has a 60 kWh battery, but as it will only be used within a city, this is enough for 14–17 hours worth of driving. It can also be charged in as little as 30 minutes.
But, it is the car’s self-driving features and approach that make it stand out.
You see, one of Tesla’s biggest advantages is its ability to gather data. It has millions of vehicles driving around with cameras recording the world around them and sending the data back to Tesla. This means Tesla has ample data to train its self-driving AI on, enabling it to perform way better than any of its rivals. But Rimac/Verne has no way of gathering data. What’s more, Tesla is so confident it can make its self-driving AI incredibly accurate that it has removed all sensors from its car, leaving it with only cameras to sense the world around it. This means it cannot catch errors or misreads by checking data from multiple sources.
So, how has Verne solved these problems?
Well, first of all, they have packed the car with loads of sensors! It has three long-distance lidar sensors, six short-distance lidar sensors, 13 cameras, and an assortment of radar sensors too. By comparison, a Model 3 only has nine cameras to sense the world around them.
Verne then uses a third-party self-driving chip and program from Mobileye. This system technically runs two self-driving programs, one making decisions purely based on the camera data and one from the other sensors. The system then checks the results from these systems against each other. This makes the system way more reliable. If the camera system misreads, its AI fails to recognise issues or suffers from AI hallucinations (where the AI imagines things that aren’t there), the system can ignore its incorrect driving orders and instead listen to the lidar/radar system. This also means the AI doesn’t have to be as accurate as Tesla’s, as the higher fidelity and highly redundant sensor system help the AI to be more accurate and can mitigate AI faults. This, in turn, means that AI needs less development than Tesla’s.
Mobileye’s system is also currently used by VW, BMW, Ford and Volvo/Geely/Zeekr, and, like Tesla, gathers data while these cars are driving. As such, Mobileye also has plenty of data on which to train its AI; in turn, Verne’s lack of driving data isn’t a problem.
There is also a difference between Tesla’s and Verne’s approach to developing this AI. Tesla appears to be making an AI that can react to road conditions in a way which means it can be used on any road, even if the AI hasn’t been trained on it before. In contrast, Verne will geo-fence their robotaxis to the city they were made for and will intensively map the city’s roads in 3D and develop a city-specific AI for said city. This constrained, highly detailed approach means Verne’s AI has to deal with far fewer variables than Tesla’s. This means Verne’s/Mobileye’s self-driving AI can be far more reliable from far less training or development than Tesla’s. As training AI is incredibly expensive and time-consuming, this is a huge advantage!
This constrained approach, combined with Verne’s vast sensor array and its inherent redundancy, will enable it to reach Level 4 autonomy years before Tesla’s unconstrained, low sensor count approach.
But why can’t Mobileye’s other customers do the same and overtake Verne?
Well, unlike every other robotaxi/self-driving car setup, Verne has actually thought about how a robotaxi business will function and has optimised it.
Verne doesn’t want to replace public transport but complement it by conducting last-mile trips instead. As such, Verne’s cars are pitched as being slightly cheaper than a taxi yet better quality and more convenient than owning a second car. They can do this as only a small fleet of cars, around 400, can replace a medium-sized city’s entire taxi fleet. These fleets will run out of “motherships”, where the cars will return to be cleaned and recharged as and when they need. In fact, the cars have internal cameras and even a digital nose to tell when the car needs cleaning, ensuring every customer has a clean, fresh-smelling Verne every time. The first Mothership and Verne service is scheduled to be set up in the Croatian city of Zagreb by 2026, with more being built in the UK and Germany by 2027.
You heard me right. I could catch a Verne to the train station in just four years!
Compare this to Tesla’s approach. They aren’t building motherships, and their cars don’t have things like digital noses. Until recently, Tesla’s plan was for Tesla owners to rent their cars out as robotaxis, meaning the cleaning and charging jobs would fall upon the owner. As someone who has previously managed an Airbnb, I can tell you this just turns you into an underpaid, glorified cleaner. It’s no wonder Tesla is now talking about launching a dedicated robotaxi platform and Tesla-owned robotaxi network, as this in-house approach has a far better business model than Tesla’s previous customer-owned one did.
So, why will Verne’s robotaxi do better than Tesla’s soon-to-be-launched dedicated robotaxi? Simply put, Verne is already in discussions with 11 cities about rolling out their services. It also seems like Verne’s detail-orientated, locally adapted approach is significantly soothing the reservations of local policymakers. Meanwhile, Tesla seems to be actively not working with lawmakers, and all the evidence points to Tesla still using its more general camera-based system for the robotaxi.
In short, Verne has not only solved the problem of making a self-driving robotaxi, but it has also actively collaborated with lawmakers and industry experts to ensure they can reach the market first and has solved the problems of how a robotaxi service will function and be deployed successfully. Meanwhile, Tesla is greedily trying to do everything in-house and is trying to bulldoze lawmakers to get their way. From a business, technological and ethical point of view, Verne absolutely demolishes Tesla.
That is why Verne is my bet for the first company to crack robotaxis.
Thanks for reading! Content like this doesn’t happen without your support. So, if you want to see more like this, don’t forget to Subscribe and help get the word out by hitting the share button below.
Sources: The Verge, Car Wow, Autocar, The Street