Chris Urmson: Self-Driving Cars at Aurora, Google, CMU, and DARPA | Lex Fridman Podcast #28 | Summary and Q&A

TL;DR

Chris Urmson, CEO of Aurora Innovation, discusses his experience with autonomous vehicles, the challenges involved, and the importance of building trust in the technology.

Key Insights

• 😣 Believing in the potential of autonomous driving and persevering through challenges are crucial in developing the technology.
• 🍁 Developing accurate maps and creating robust hardware and software systems are essential for autonomous vehicle success.
• ❓ Effective leadership involves empowering others and recognizing their potential.
• 🪡 Autonomous driving technology needs to be demonstrated as safe, reliable, and superior to human performance to gain public trust.
• 🚙 Lidar, cameras, and radars are all important sensors for autonomous vehicles, contributing to a comprehensive perception system.
• 🧑‍🏭 Level 2 autonomous systems require careful marketing and human factor considerations to prevent over-reliance on the technology.
• 🚴 Achieving full autonomy requires solving complex technical challenges, such as predicting the behavior of pedestrians and cyclists.

Transcript

the following is a conversation with Chris Urmson he was a CTO of the Google self-driving car team a key engineer and leader behind the Carnegie Mellon University autonomous vehicle entries in the DARPA Grand Challenges and the winner of the DARPA urban challenge today he's the CEO of Aurora innovation and the autonomous vehicle software company he... Read More

Questions & Answers

Q: What were the major challenges faced during the DARPA Grand Challenge and the DARPA urban challenge?

The major challenge was believing that autonomous driving was possible and overcoming the technical difficulties such as mechanical issues, sensor limitations, and developing accurate maps.

Q: Is lidar an essential part of autonomous vehicle technology?

Yes, lidar is essential, but it should be used in combination with cameras and radars. The different sensors provide a holistic view of the surroundings, improving the robustness of the autonomous system.

Q: Will level 2 autonomous vehicles be safe without over-trusting the technology?

No, because humans tend to trust technology after positive experiences, even if they don't fully understand its limitations. It is important to educate the public about the capabilities of level 2 systems and always prioritize human safety.

Q: How do we demonstrate that autonomous vehicles are safe?

Demonstrating safety requires a combination of diligent engineering work, evidence of system capabilities, and gaining the trust of regulatory bodies. Metrics should be developed to showcase the performance of autonomous systems compared to human drivers.

Q: What were the major challenges faced during the DARPA Grand Challenge and the DARPA urban challenge?

The major challenge was believing that autonomous driving was possible and overcoming the technical difficulties such as mechanical issues, sensor limitations, and developing accurate maps.

More Insights

• Believing in the potential of autonomous driving and persevering through challenges are crucial in developing the technology.

• Developing accurate maps and creating robust hardware and software systems are essential for autonomous vehicle success.

• Effective leadership involves empowering others and recognizing their potential.

• Autonomous driving technology needs to be demonstrated as safe, reliable, and superior to human performance to gain public trust.

• Lidar, cameras, and radars are all important sensors for autonomous vehicles, contributing to a comprehensive perception system.

• Level 2 autonomous systems require careful marketing and human factor considerations to prevent over-reliance on the technology.

• Achieving full autonomy requires solving complex technical challenges, such as predicting the behavior of pedestrians and cyclists.

• Large-scale deployment of autonomous vehicles is likely within the next decade, starting in urban and suburban environments.

Summary:

In this interview with Chris Urmson, CEO of Aurora Innovation, he discusses his experiences in the field of autonomous driving and the challenges that come with it. From his participation in the DARPA Grand Challenge and DARPA Urban Challenge to his thoughts on leadership and the future of autonomous vehicles, Urmson provides insightful answers to a wide range of questions.

Questions & Answers:

Q: What technical or philosophical things did you learn from the DARPA Grand Challenge and DARPA Urban Challenge?

Urmson states that the main takeaway from these challenges was the realization that autonomous driving was indeed possible. Despite the initial difficulties, the teams were able to overcome obstacles and make it happen.

Q: At what point did you personally believe autonomous driving was possible?

Urmson explains that from the beginning, the team had the belief that it could be done. While they didn't know the exact path to success, they were determined to try different approaches and make it happen.

Q: What were the biggest pain points in the technical evolution of autonomous vehicle systems?

Urmson mentions several pain points including mechanical issues, sensor technologies, hardware and software challenges, algorithms for mapping and localization, as well as perception and control.

Q: What were the major shifts in autonomous vehicle technology from the first Grand Challenge to the urban challenge and to today?

Urmson highlights the importance of HD mapping in the first Grand Challenge and multi-beam lidar technology in the urban challenge. He also mentions advancements in Bayesian estimation techniques and the use of sensors like lidar, cameras, and radar for perception and localization.

Q: What did you learn about leadership from reading about Red Whittaker, the leader of the DARPA Urban Challenge team at CMU?

Urmson explains that he learned the importance of tackling difficult challenges and empowering others to be leaders. Red Whittaker saw the potential in his team members, even undergraduates and graduate students, and trusted them to take on responsibilities and achieve great things.

Q: How do you foresee the deployment and scaling of autonomous vehicles in the future?

Urmson believes that within ten years, we will see a large-scale deployment of autonomous vehicles, with thousands of them on the roads. He envisions the initial deployment in urban and suburban environments and expects a gradual scaling process.

Q: How do we demonstrate the safety of autonomous vehicles to the world?

Urmson emphasizes the importance of a thorough and diligent approach in engineering autonomous systems, including a functional safety process. He mentions that a combination of evidence from simulations, testing, and on-road data can be used to demonstrate the capabilities and safety of the systems.

Q: Can a metric be created to demonstrate safety outside of fatalities?

Urmson believes that metrics related to human performance in different tasks, such as detecting traffic lights and making left turns, can be used to demonstrate safety. Comparing the failure rates of the autonomous system to those of humans can provide a compelling story about the system's capabilities.

Q: How do you win the hearts and minds of people to accept autonomous vehicles as a part of their lives?

Urmson suggests letting people experience autonomous vehicles firsthand. Skepticism and doubt are natural, but once people see that the technology is safe and useful in their daily lives, they will start to accept it. Creating a compelling and valuable user experience is key to winning people's trust and adoption.

Q: Are there any breakthroughs that could accelerate the deployment of autonomous vehicles?

Urmson mentions that a breakthrough in the capability to accurately perceive and forecast the surrounding environment could greatly accelerate the deployment of autonomous vehicles. This would involve having a perfect model of what has happened, is happening, and will happen around the vehicle.

Summary & Key Takeaways

• Chris Urmson reflects on his involvement in the DARPA Grand Challenge and the DARPA urban challenge, highlighting the importance of the belief that autonomous driving was possible and the value of perseverance.

• He discusses the key pain points in developing autonomous vehicle technology, including mechanical, sensor, hardware, software, and algorithmic challenges.

• Urmson shares valuable insights on leadership, emphasizing the importance of taking on difficult tasks and seeing the potential in others.