Here are the questions we collected and answers replied by the presenter: Q1. How much computing power (TOPS) does this project need to achieve? It depends on the task. For driving scene perception, we need algorithms that run in real time (e.g., milliseconds). For vehicle positioning, near-real time performance (e.g., 1 second) is acceptable. For the generation of HD maps computational efficiency is not very critical since the process can be done offline. Q2. Nowadays, more and more models will be equipped with Laser Radar, and the price of Laser Radar is getting lower. Do you think Laser Radar (complemented by millimeter wave radar, high-definition camera, and high-precision GPS) is the ultimate solution for L4-L5 autonomous driving in the future? Yes, I believe L4-L5 automation cannot be achieved using a single sensor and a multi-sensor approach integrating lidar, camera, GNSS, and radar is needed.
Here are the questions we collected and answers replied by the presenter:
Q1. How much computing power (TOPS) does this project need to achieve?
It depends on the task. For driving scene perception, we need algorithms that run in real time (e.g., milliseconds). For vehicle positioning, near-real time performance (e.g., 1 second) is acceptable. For the generation of HD maps computational efficiency is not very critical since the process can be done offline.
Q2. Nowadays, more and more models will be equipped with Laser Radar, and the price of Laser Radar is getting lower. Do you think Laser Radar (complemented by millimeter wave radar, high-definition camera, and high-precision GPS) is the ultimate solution for L4-L5 autonomous driving in the future?
Yes, I believe L4-L5 automation cannot be achieved using a single sensor and a multi-sensor approach integrating lidar, camera, GNSS, and radar is needed.