BIG Data in Tesla Inc.
As a car company that is also categorized as a “big tech” company, Tesla not only revolutionized the car industry with its EVs (Electric Vehicles) but also with its usage of Big Data. We would dive deep into the core technology behind Tesla, specifically how Tesla acquires data and uses such data to train safety and self-driving models on their cars. We would also discuss how Tesla gives a customized driving experience that resulted in the top customer rating in the industry.
Big data for self-driving
Impressions for Tesla cars? “Self-driving” would be many’s first answers. Not only was Tesla the biggest player in the electric vehicle world, but it was also one of the pioneers and leaders in the field of autonomous driving. Tesla provides level 2 self-driving as an optional feature on all models. It is sold to around 20-40% of all Tesla buyers globally with $5-10 billion in sales, a big chunk of the global Advanced Driver Assistance Systems Market size of $18.2 billion in 2022 [1].
Tesla achieves their self-driving features through AI/Machine Learning combined with an incredible system of data acquisition and aggregation from various sources. All Tesla models are equipped with sets of sensors, cameras, and radars, that constantly collect data from individual cars’ detailed surroundings from the outside and driver and passenger gestures within the car. In addition to the data generated within the cars, the cloud server would also collect real-time information about the car such as information on GPS, traffic, weather, object maps (humans, animals, traffic signs, etc.), and other connected cars (see Figure 1 as a demonstration) [2].
A positive virtuous circle also exists, as the cars run on self-driving models would give more data that could be fed back into the model to provide better self-driving capability for the cars. As more and more Tesla cars are driven on the road, Tesla designs a network of “fleet learning,” which refers to the fact that the whole population of Tesla cars could learn from all individual’s newly acquired data, to enhance its data collection and analytic process [2]. All Tesla models could be updated remotely through an internet connection just like the small phones, a feature they named “over-the-air” software updates [3]. Thus, any newly improved self-driving models could be implemented on a mass scale within a short period, giving chances for more iterations of improvements.
Of course, the self-driving model fed on big data is far from perfect. The data collected from faulty sensors could be noisy or untrue. Customer feedback from self-driving could also be subjected to biases within the drivers. The large amount of data generated from each car (which includes large-size video data) could pose significant burdens to data cleaning and data analysis that require high computational power not only on the cloud server but also within individual cars.
Big data for safety and a customized experience
Tesla models come standard with a set of safety avoidance features that they called “active safety features” [4]. Tesla cars, with the help of AI/ML models, are able to use their various sensors to identify road hazards, such as icy roads, and notify the drivers about the underlying dangers. Potential collisions predicted by the cars’ advanced custom-made computers would also be forecasted to the drivers for immediate action to mitigate risks [5]. In the emergency case when the drivers are unable to respond in time to avoid collisions, Tesla cars would perform actions automatically to avoid such incidents with features such as “Automatic Emergency Braking,” “Lane Departure Avoidance,” and “Emergency Lane Departure Avoidance” [4].
All these intelligent safety features require accurate models or algorithms that would not only identify the risks from various forms of data in real-time with high precision and accuracy but also derive the best mitigation plans accordingly. Tesla officially said that they collected data “from millions of Tesla vehicles to learn how collisions happen– and how we (Tesla) can help prevent them in the future” [3].
However, similar problems of using big data for ML prediction also exist. Such predictions would have to be in the balance of being conservative vs. being liberal to still be functional, given the context of safety.
Big data for customized driving experience
According to a 2021 report, Tesla has the highest customer satisfaction rate out of most car makers [6]. Tesla enhances its big data strategy to create personalized experiences geared toward each individual customer. With the use of various external or internal sensors, each Tesla vehicle is able to provide a customized driving experience such as the songs played upon starting the car, the mirror positions, air-conditioning volumes, navigation paths designed for the specific time of day, etc. Each car is able to learn from its owners and tailor to the owners’ needs. The dynamic personalization feature of Tesla cars is even able to distinguish different drivers in the car and thus provide a customized experience accordingly [7]. Finally, Tesla’s analytic team obtained data such as text from online forums to find demands and complaints about their vehicles as bases for future improvements [8].
Citations
[1] Brian Wang, “Tesla Full Self Driving, Autopilot and the Driver Assist Market | NextBigFuture.Com,” June 26, 2023, https://www.nextbigfuture.com/2023/06/tesla-full-self-driving-autopilot-and-the-driver-assist-market.html.
[2] Bipin Karki, “Big Data and Analytics in Tesla Inc.,” January 18, 2020, https://www.linkedin.com/pulse/big-data-analytics-tesla-inc-bipin-karki/.
[3] “Safety,” Tesla, accessed October 17, 2023, https://www.tesla.com/safety.
[4] “Model 3 Earns 5-Star Safety Rating from Euro NCAP,” Tesla, accessed October 17, 2023, https://www.tesla.com/blog/model-3-earns-5-star-safety-rating-euro-ncap.
[5] Vaishnavi Yada (Amira), “How Tesla Is Using AI and Big Data Analytics in Their Self Driving Cars?,” Dare To Be Better (blog), February 16, 2023, https://medium.com/dare-to-be-better/how-tesla-is-using-ai-and-big-data-analytics-in-their-self-driving-cars-7072e410c1b8.
[6] Jeremy Korzeniewski, “Tesla Leads and Infiniti Bleeds in Consumer Reports’ Satisfaction Survey,” Autoblog, accessed October 17, 2023, https://www.autoblog.com/2021/02/08/car-truck-owner-satisfaction-survey-consumer-reports/.
[7] Blake Morgan, “3 Ways Tesla Creates A Personalized Customer Experience,” Forbes, accessed October 17, 2023, https://www.forbes.com/sites/blakemorgan/2021/05/10/3-ways-tesla-creates-a-personalized-customer-experience/.
[8] Vikram Singh, “Tesla: A Data Driven Future,” accessed October 17, 2023, https://d3.harvard.edu/platform-digit/submission/tesla-a-data-driven-future/.
[9] Ria Cheruvu, Big Data Applications in Self-Driving Cars, 2015, https://doi.org/10.13140/RG.2.2.12302.05445.
Thanks for your work, Sam! We’ve learned the mechanism of self-driving car from DeepMap. The blog introduces some similar features and some different features of Tesla. I’m curious about whether there is any key difference between the self-driving mechanism and the use of big data between DeepMap and Tesla.
Besides, you mentioned that “Potential collisions predicted by the cars’ advanced custom-made computers”. I assume there is a real-time machine learning predictor inside the car system, but I feel interested to learn that the car can act like a computer. Does it use chips or built-in systems for the ML/AI process? For the real-time traffic prediction machine learning model, is there any difference from the traditional machine learning models we use?
Hi Sam, Thanks for your post. The safety features is definitely one of the key discussed points of AV’s . They have been several reports and discussions regarding this. It would be interesting to understand or see key statistics on how “Automatic Emergency Braking,” “Lane Departure Avoidance,” and “Emergency Lane Departure Avoidance has reduced fatalities.
Wow Sam, great minds think alike!
I think one thing I’ve always found interesting about Tesla is their reluctance to use LiDAR devices to augment their sensor offerings. Historically, they’ve made the argument that it is unnecessary on technical grounds (i.e. their cameras and other sensors work just as well), but I’ve always believed that it was because of cost. But as the prices of LiDAR decrease steeply (not unlike microprocessors in the past decades), I’m curious if Tesla will reverse course and adopt it. LiDAR is still regarded as a superior technology to traditional cameras, and as regulatory scrutiny increases, it would almost be irresponsible of Tesla not to at least offer it to quell public concerns.