Aurora CEO Chris Urmson on building the roadmap to fully autonomous vehicles | E1228

Episode Summary

Executive Summary: Chris Urmson, CEO of Aurora, discusses his journey from DARPA challenges to leading a self-driving truck company. He explains the evolution of autonomous vehicle technology, Aurora's focus on trucking for scalability, the three-pillar safety case, and the business model of providing the Aurora Driver as a service. He addresses challenges like false positives/negatives, the importance of simulation, and the need for multiple sensor modalities. The conversation highlights the immense opportunity in freight and the path to commercial deployment.

Main Topics: History of Self-Driving (DARPA Challenges) (Priority: 5/5): Urmson recounts his involvement in the DARPA Grand Challenges from 2003-2007, detailing the early struggles and breakthroughs in autonomous vehicle technology, including the 2003 race where his team's vehicle went only 7.5 miles out of 150, and the 2005 race where five vehicles finished. Technology Evolution (Hardware/Software) (Priority: 4/5): Discussion of the dramatic improvements in sensors (from 1MP cameras and single-beam LIDAR to multi-beam LIDAR, radar, and high-res cameras) and computing (from Itanium processors to NVIDIA GPUs). Urmson notes that software is now the bigger challenge. Aurora's Strategy (Trucking First, Partnerships) (Priority: 5/5): Aurora is focusing on trucking as the first product, starting in Texas, because freeway networks are self-similar and customer expectations are less demanding. They partner with PACCAR, Volvo, Toyota, and Uber rather than building everything in-house. Safety Approach (Three Pillars, False Positives/Negatives) (Priority: 5/5): Urmson outlines Aurora's safety case: organizational/operational safety, functional safety (graceful failure), and safety of the intended function (SOTIF). He explains the false positive/negative problem and why multiple sensor modalities are critical to avoid failures like the Tesla crash. Business Model (Driver as a Service, Data) (Priority: 4/5): Aurora plans to provide the Aurora Driver as a service, charging per mile. They use simulation for efficient development and targeted real-world testing, rather than gathering millions of miles of data. The cost of equipment is expected to drop from hundreds of thousands to $10-20k. Future Outlook (Consolidation, Timeline) (Priority: 3/5): Urmson predicts consolidation to a handful of players. He believes self-driving trucking will deploy in the coming years, with robo-taxis following later. The U.S. freight market is $700 billion, and ground transportation is $2-3 trillion, offering huge opportunities.

Key Arguments: Self-driving technology has evolved dramatically from DARPA days; software is now the bigger challenge. Trucking is the right first application due to self-similar freeway networks and less demanding customer expectations. Safety requires a comprehensive case covering organizational, functional, and intended function safety. Multiple sensor modalities (LIDAR, radar, camera) are essential to avoid failures like the Tesla crash. Simulation is a key tool for efficient development and validation, reducing the need for massive real-world data collection. The industry will consolidate to a handful of players; Aurora aims to be one through partnerships and focus.

Data Points: DARPA 2003 distance: 7.5 miles out of 150 - First Grand Challenge: only 5% of course completed. DARPA 2005 finishers: 5 vehicles - Second Grand Challenge: five vehicles completed the 150-mile desert course. Aurora LIDAR range: Several hundred meters (over 1000 feet) - First Light LIDAR enables long-range detection and velocity measurement. U.S. freight market size: $700 billion - Freight in the U.S. alone is a $700 billion business. U.S. ground transportation market: $2-3 trillion - Total ground transportation in the U.S. is $2-3 trillion. Current prototype equipment cost: Hundreds of thousands of dollars - Small-volume prototype equipment costs $100k-$200k. Target equipment cost: $10,000 - $20,000 - Expected cost at scale for the Aurora Driver system. Car utilization rate: 4% - Personal cars are driven only about 4% of the time they are owned.

Pivotal Quotes: "We have been making fundamental advances ourselves, but we've also been benefiting from advances in the rest of industry." — Chris Urmson: On the evolution of self-driving technology and the role of broader tech advancements like smartphones. "The safety benefits come along for the ride... you use it because of all these other advantages." — Chris Urmson: Explaining why self-driving vehicles will be adopted for convenience, with safety as a side benefit. "Don't hit the stuff that's there. It's pretty straightforward." — Chris Urmson: On dealing with the long-tail problem of rare objects; the system should avoid hitting anything it detects.

Implications: Self-driving trucking will likely deploy in Texas within a few years, starting with hub-to-hub freeway routes. The technology will save lives and reduce costs, but requires careful safety validation. The industry will consolidate, and Aurora is well-positioned with its partnerships and focus on trucking. Consumers may see robo-taxis later, but trucking is the near-term commercial opportunity.