Where Are All the Autonomous Vehicles?

🌈 Abstract

The article discusses the current state of autonomous vehicles (AVs) and the challenges that are delaying their widespread adoption. It outlines three key reasons why driverless cars are taking longer than expected to become a reality:

1. The bar for safety is higher than expected, with autonomous vehicles needing to pass rigorous safety cases and address a vast number of edge cases.
2. Scaling driverless cars is harder than anticipated, requiring significant infrastructure and support systems beyond just the vehicles themselves.
3. The economics of driverless cars are shaky, with the costs of the required technology and support systems outweighing the projected revenue benefits.

The article also touches on the progress of Tesla and other automakers with their automated driving features, noting that they are still far from true self-driving capabilities.

🙋 Q&A

[01] The bar for safety is higher than we thought

1. What is a "safety case" in the context of autonomous vehicles? A safety case is how an AV company gains its own conviction about the safety of its vehicles in order to avoid liability issues. It involves identifying and addressing a vast number of potential edge cases and scenarios that could lead to accidents or injuries.

2. Why is the "long tail" problem a challenge for autonomous vehicle safety? The "long tail" problem refers to the countless uncommon but important edge cases that AVs need to be able to handle safely. There are nearly infinite scenarios to test for, making it extremely difficult to ensure the vehicle can navigate all possible situations.

3. What is the "operational design domain (ODD)" and how does it relate to safety? The ODD refers to the specific context in which an AV can safely operate, such as certain roads, weather conditions, and other constraints. Limiting the ODD helps reduce the number of scenarios that need to be tested, but the complexity of operating within even a narrow ODD remains high.

4. What is the "absence of unreasonable risk (AUR)" and why is it difficult to determine for autonomous vehicles? AUR is the safety threshold that autonomous vehicles need to meet, based on "societal moral concepts." It's unclear whether AVs need to be better than the average human driver, better than the best human driver, or something else entirely to be considered "safe enough."

[02] Scaling driverless cars is harder than we thought

1. What are some of the key infrastructure requirements for scaling a fleet of autonomous vehicles? Key requirements include high-definition mapping, remote support/tele-assist teams, in-field support for vehicle recovery, charging infrastructure, maintenance facilities, spare parts supply chains, cleaning facilities, and insurance solutions.

2. Why is the transfer of AV technology and capabilities between different geographic locations a challenge? Even if an AV system is proven to work well in one city, factors like vehicle/pedestrian behaviors, road configurations, and environmental conditions can vary significantly between locations, requiring the entire safety case process to be repeated.

3. How do the labor and support costs for a fleet of robotaxis compare to traditional ride-sharing services? The support labor costs for robotaxis are higher, as the human-to-vehicle ratio is currently around 1.5 humans per vehicle to handle remote assistance, in-field operations, and other support functions.

[03] The economics of driverless cars are shaky

1. What are some of the key cost factors that make the economics of robotaxi services challenging? Costs include the software platform, high-priced AV hardware (around $150-200k per vehicle), maintenance of specialized sensors and components, the labor-intensive support teams, HD mapping expenses, and the need for technology upgrades over time.

2. How do the revenue projections for robotaxi services compare to the actual costs? While the revenue potential looks promising based on ride-sharing models, the actual costs of the required technology and support infrastructure are much higher than originally anticipated, making it difficult to achieve profitability.

3. What is the current funding situation for autonomous vehicle companies, and why are investors becoming more cautious? Autonomous vehicle startups have required significant venture capital funding to develop their technologies, but investors are now becoming more hesitant to pour in additional billions as the path to profitability remains elusive.