Tesla, Waymo, & Lyft: Hurdles to Autonomous Driving
Aug 28, 2020
auto_awesome
A former VP at Google, now EVP at Lyft, discusses the challenges and developments in autonomous driving. Topics include enhancing 3D mapping with LIDAR and radar data, detailed breakdown of the AV stack, the role of proprietary data, evolution of perception, various business models in the industry, and the use of LIDAR in autonomous driving.
Lyft Level 5 focuses on tailored self-driving cars for rideshare, emphasizing unique service needs.
Google Street View started as a vision project using Lidar for smooth transitions between bubbles.
Different companies explore varied AV business models, with slow deployment expected due to challenges.
Deep dives
Lyft's Autonomous Driving Focus on Rideshare
Lyft's team, Lyft Level 5, is dedicated to building self-driving cars specifically for rideshare. The team was founded about three and a half years ago and has been working on developing autonomous technology for Lyft. This focus allows Lyft to tailor the technology to meet the unique needs of rideshare services.
Fascinating Origins of Google's Street View
Google started its Street View project as a collaboration between Stanford and Google, driven by Larry Page's vision to bring the world to people through collecting street-level imagery. The project was initially a 20% project at Google, but it gained traction and support, leading to the creation of an end-to-end demo that showcased the potential of this technology. From there, Google Maps began utilizing imagery and deriving knowledge from various sources to improve maps.
The Role of Lidar in Street View
Lidar played a crucial role in Google's Street View project, particularly in creating a smooth transition between street view bubbles. By using Lidar, Google was able to project imagery onto a coarse 3D model and achieve a seamless transition from one bubble to another, enhancing the user experience. While imagery was the primary focus, Lidar also provided valuable information for 3D understanding of the environment.
The Business Models in Autonomous Vehicles
In the autonomous vehicle space, different companies are exploring various business models. Some, like Waymo and Cruise, aim to build and deploy self-driving fleets for rideshare and taxi services. Others, like Tesla and traditional OEMs, are working towards selling fully autonomous vehicles directly to consumers. There are also companies focusing on providing AV maps to the industry, as HD maps are crucial for safe autonomous navigation. Partnerships between tech companies and OEMs are common as both sides bring unique expertise and resources to the table.
Challenges and Timelines for Autonomous Vehicles
The deployment of autonomous vehicles will likely take longer than initially anticipated. While small-scale deployments are expected in the next couple of years, widespread commercialization will take additional time. Challenges include developing robust planning systems, addressing system failures, and proving safety in real-world scenarios. Timelines may also depend on potential breakthroughs in AI that enable machines to better extrapolate and handle new situations. Despite these challenges, the trend towards electric and autonomous vehicles is expected to continue, although it may take several decades to fully mature.
