Content area
Full Text
Tremendous efforts have been made in the field of vehicle automation. That is, developing advanced sensors and algorithms that are installed on vehicles to imitate or eventually transcend the complexity of human drivers. This approach focuses solely on vehicle-based technologies and takes the road as the status quo. Many players are investing in this vehicle-based approach, from giant high-tech companies and car manufacturers, to small start-ups, by taking advantage of the latest development in hardware, software, and communication technologies. Despite the significant progress that has been made, automated vehicles are still not safe and reliable enough for large scale deployment, as indicated by incidents during automated vehicle testing.1,2
Self-Driving Cars: A Vehicle-based Approach for Automated Driving
According to the 2018 Gartner Hype Cycle for Emerging Technologies report, automated driving still has more than 10 years to reach the "Plateau of Productivity."3 Although Waymo just launched the first commercial self-driving taxi service, Waymo CEO John Krafcik said, "Self-driving cars will require driver assistance for many years to come," and "the technology is 'really, really hard.'"4 The CEO of the Toyota Research Institute Gill Pratt said at the 2019 Consumer Electronics Show, "Now, none of us in the automobile or IT industries are close to fully answering these questions."5
The major challenges for automated vehicles include: a) sensing the environment, b) understanding human intents, and c) the resulting high cost. Sensing, or situation awareness, is a critical component for automated driving. A typical automated vehicle relies on a range of different sensors, including but not limited to LiDAR (light detection and ranging), radar, cameras, and ultrasonic. Those sensors are used to cover different ranges to detect the environment and moving objectives about the vehicle. Advanced artificial intelligence (AI) algorithms are developed and deployed on the vehicle to make sense of the data streams from those sensors. The AI algorithms are also supposed to learn from the human drivers' behavior and figure out how to drive the vehicle and how to interact with everything around it. What makes it more complicated is that automated vehicles also need to learn how to interact with each other, especially among vehicles from different manufacturers.
All those sensors and software/hardware are making the cost of self-driving vehicles currently beyond the affordable range...