[Fig. 1] Diagram of Tesla’s NoA (Navigation on Autopilot): Automatic Lane Changing based on Navigation Route & Traffic Flow (Source: CNN Business)
Being the first long-range electric vehicle (EV) in the market, Tesla is considered the ‘true innovator’ in the automotive industry. From its battery pack technology, Tesla has ‘innovated’ the travel range of EVs, which used to be ephemeral compared to vehicles with an internal combustion engine. Tesla’s ‘supercharger’ technology alleviated the customers’ burden of purchasing EVs by lowering their maintenance costs using simple powertrain architecture—allowing Tesla’s sales to soar. Ultimately, Tesla Model S was awarded as Motor Trend’s “Best Car of the Year”. Tesla’s success motivated other premium auto brands, such as Porsche, Mercedes-Benz, and BMW, to develop and release their own electric cars. In fact, automotive journalists consider Tesla’s existence brought the advent of electric cars ten years earlier than it was supposed to.
Apart from bringing innovation to the electric vehicle industry, Tesla has brought significant innovation in another automotive aspect: self-driving cars. Since its release in October 2014, Tesla Autopilot was the first driver-assistance driving system that converges to the level of self-driving available to the public.
As alluded to in its name, Autopilot is known to provide both steering and acceleration/deceleration assistance. However, there are similar driver assistance packages available from other premium automotive manufacturers, such as Distronic Plus from Mercedes, Driving Assistance Professional from BMW, or Audi AI.
However, among such services, why is Autopilot considered a self-driving system? What differentiates Tesla from the others? Everything, including the manufacturer’s effort in the development of self-driving technology.
In hardware facets, the number of cameras is the key difference between Tesla’s Autopilot and driving assistance/self-driving provided by other manufacturers. For instance, Mercedes Benz only relies on three front-view cameras installed at the rearview mirror and radar at the front emblem/grill (The glass-like board is the radar). In contrast, Tesla cars with Full Self-Driving Capability are equipped with 8 cameras that provide 360 degrees of visibility at up to 250 meters of range. These cameras installed allow Tesla’s autopilot to drive like, or even better than, humans since the computer is aware of the complete surroundings of the car.
[Fig. 2] Diagram of Tesla Cameras for Autopilot (Source: Electrek)
Besides the number of cameras, another key difference between Tesla and other cars is the use of wireless networks. While other cars use a wireless technology called TPEG only for checking traffic congestion status, Tesla’s cars are equipped with an on-board computer system capable of LTE connection, allowing the owners to ‘update’ their cars to Tesla’s newest releases (similar to software updates in phones or computers). The updates include not only minor changes such as software UI but also major changes such as the car’s suspension settings and the Autopilot system. Within each of these updates, Tesla releases a newly trained self-driving AI model to every car, which significantly enhances the self-driving accuracy.
How was Tesla able to rapidly improve its self-driving system? The answer is hidden in Tesla’s “fleet” infrastructure, a hidden communication system integrated into Tesla’s cars. In essence, when the user disables the autopilot system, the Tesla system considers it an ‘emergency situation’. Using LTE, the car sends its surroundings, driver’s steering, and pedal input to the headquarters, where the AI model training takes place. Although AI technologies are being consistently developed for better accuracy, data quantity is the most important factor in training. Due to this communication system established, Tesla can secure a lot of data for training its autopilot system, making Tesla the unbeatable leader in self-driving.
Q & A:
Jennah: How can Tesla’s autopilot system benefit the future? In a broader sense, are there any plans for the future of Tesla and/or other technological companies that incorporate the mechanism?
Tesla’s autopilot is the pioneer in developing self-driving. By selecting the self-driving hardware option, owners can receive a car with hardware capable of self-driving. Theoretically, self-driving is feasible since the cameras have a complete view of the car’s surroundings. Currently, Tesla released its self-driving beta to U.S. customers. Although it is estimated to take a significant amount of time to receive official approval from the local governments, there is no doubt that Tesla is leading the path to self-driving cars. A strong competitor against Tesla is Mercedes-Benz, which recently received approval from the German government with Level 3 self-driving. Also, Tesla’s autopilot led to the establishment of many self-driving technology startups, such as openpilot from comma.ai or Zoox. With combined efforts, complete self-driving may not be a far future.
David: How do other car companies and their systems compare to Tesla? What about other car companies with autopilot systems? Do they not have the same advanced AI systems?
When it comes to competition, Tesla’s autopilot is in an ambiguous position. It is too good to call it a “driver assistance system.” Compared to other driver-assistance systems available in the market, Tesla’s autopilot system beats the competition. Although other driving assistance systems do a proficient job in keeping the car in straight lanes, they are not capable of maintaining the car in its lane when there is no clear vision provided, such as driving around corners or driving in rain. For example, when going around highway junctions or corners, Tesla is the only system that analyzes the curve and slows down the car accordingly. Also, when clear visibility of the lanes is not provided due to heavy rain or road construction, Tesla’s AI analyzes the arrangement of other cars ahead and drives accordingly, which is still considered evolutionary in self-driving. Also, there are no manufacturers that utilize self-driving with the navigation route except Tesla’s NoA (Navigation on Autopilot) system. Tesla’s NoA, part of the software version 9.0 update for cars with the Enhanced Autopilot package, “guides a car from a highway’s on-ramp to off-ramp, including suggesting and making lane changes, navigating highway interchanges, and taking exits” (Tesla Official Blog). These are some key points that distinguish Tesla from other driver assistance systems.
Anna: Can you elaborate more on what LTE technology is and what relation it has with Tesla?
The 4G/LTE technology in Tesla is not different from that in our cell phones. Tesla uses this technology to not only send driving footage but also receive updates. Furthermore, the LTE technology allows owners to enjoy online content, such as Netflix, YouTube Spotify, or online games, while they are waiting in their car while charging.
Hannah: How is Tesla developing the autopilot system? For example, what sensors and
programs do the AI use to differentiate cars from pedestrians?
Similar to other self-driving companies, Tesla develops the autopilot system with the help of object detection, which is used in vision artificial intelligence. All driving footage of every Tesla car is sent to Tesla headquarters for the development of Autopilot. In Tesla’s Autopilot team, each item in the footage is categorized: cars, trucks, pedestrians, cyclists, etc. Andrej Karpathy, director of artificial intelligence and Autopilot Vision at Tesla, claims that Tesla has established an automated system for this categorizing process. With the labeled footage as training data, the vision AI model is trained using the computing power at the headquarters. After, the ideal parameters of the trained AI model are distributed to all Tesla cars through software updates.
Ian: What are some possible drawbacks of Tesla’s autopilot system? What are some of the safety concerns and how are these concerns addressed by Tesla's autopilot system.
“So good that it’s bad.” Despite Tesla consistently claiming Autopilot as a driver-assistance program, people tend to consider it as a self-driving system. This leads to some drivers completely relying on Autopilot to drive and not paying attention to the road. Abiding by the local regulations, Tesla always prompts drivers to grab the steering wheel; however, the car’s steering wheel recognizes the driver’s hands if a certain weight is exerted on the steering wheel. As a result, people tend to place weights on the steering wheel to avoid the following safety measures.
John: Could you explain more on how Tesla motivated other premium automobile companies to develop electric cars? For example, what are some of the electric car models from other companies and when were they developed?
When first released, Tesla Model S showed the world that electric cars can be not only luxurious but also travel long distances, reducing the burden of charging. In fact, it was the first electric car to break the 400 miles range barrier.
Being crowned as the 2013 Motor Trend’s Best Car of the Year, Tesla Model S became the benchmark of other companies, who expedited their development of premium electric cars with long ranges.
The best example is Porsche’s Taycan. Similar to Model S, the car boasts a sleek, four-door coupe design. Also, the variants available in Taycan have similar specifications to Model S. For example, Taycan Turbo S is released with a similar amount of horsepower (761 hp) and acceleration time to Tesla Model S Performance (762 hp).
All brands, such as Mercedes Benz, BMW, Audi, etc, are seeking to develop electric vehicles that best reflect the brand values. For instance, the Mercedes Benz EQS is focused on the comfort of the ride, while the BMW iX is more focused on the brand’s slogan: ‘sheer driving pleasure'.
Melissa: What is the public’s response/reaction to Tesla’s autopilot system? Are there any feedbacks that Tesla should consider while further developing their system?
The public’s opinion is quite polarized regarding Tesla’s Autopilot system. Looking from the consumer’s perspective, Tesla’s autopilot system has far better accuracy than other services and includes enhanced driving skills, such as slowing down before corners and inferring where to go by looking at other cars when the lanes can’t be detected due to weather or construction. Due to these benefits, Tesla owners praise Tesla’s autopilot system. Also, since Tesla makes an approach similar to IT companies, the IT industry also praises Tesla’s autopilot system.
Although Tesla emphasizes that autopilot is a driver ‘assistance’ program, opposers argue that Tesla is scamming the consumers by stating that autopilot is a self-driving system (as mentioned above). On the flip side, it can be considered that Tesla’s autopilot is quite accurate that some confuse it as a self-driving system.
Tesla is quite open to feedback. Although there are no sales agents for each customer in Tesla’s sales infrastructure, customers’ feedback is well contributed through updates. Some tend to send their complaints directly to Elon Musk, the CEO, through Twitter. Also, since Tesla minimizes every process like an IT company, it can quickly embrace changes.
Sources:
“Andrej Karpathy - AI for Full-Self Driving at Tesla.” YouTube, 20 Apr. 2020, https://youtu.be/hx7BXih7zx8.
MacKenzie, Angus. “Mercedes-Benz Drive Pilot Autonomous First ‘Drive’: We Try a World's First Driverless System.” MotorTrend, MotorTrend, 6 May 2022, https://www.motortrend.com/news/mercedes-benz-drive-pilot-eqs-autonomous-driverless-first-drive-review/.
Matousek, Mark. “Motortrend Named Tesla's 2013 Model s Its 'Ultimate' Car of the Year among the Vehicles That Have Won the Honor in the Past 7 Decades.” Business Insider, Business Insider, 11 July 2019, https://www.businessinsider.com/tesla-model-s-named-motortrend-ultimate-car-of-the-year-2019-7.
Comments