数据大用途：训练自动驾驶汽车

2020-12-28何高伦

疯狂英语·读写版 2020年12期

关键词：费力比较级句式

何高伦

为了安全起见，自动驾驶汽车必须准确跟踪行人、非机动车和周围其他车辆的运动轨迹。一般而言，可用于训练跟踪系统的道路和交通数据越多，结果越好。

难词探意

1. autonomous /???t?n?m?s/ adj. 自治的;自主的;自发的

2. navigate /?n?v?ɡe?t/ v. 导航;航行

3. laborious /l??b??ri?s/ adj. 辛苦的;耗时费力的

4. reverse /r??v??s/ v. 使反转

Generally speaking， the more road and traffic data available for training tracking systems， the better the results will be. Researchers have found a way to unlock a mountain of autonomous driving data for this purpose. “Our method is much more precise than previous methods because we can train on much larger datasets，” said Himangi Mittal， a researcher in CMUs Robotics Institute.

Most autonomous vehicles navigate primarily based on a sensor called lidar， a laser device that generates （生成） 3D information about the world surrounding the car. This 3D information isnt images， but a cloud of points. One way the vehicle makes sense of this data is using a technique known as scene flow. This involves calculating the speed of each 3D point.

In the past， state?of?the?art methods for training such a system have required the use of labeled datasets， which is laborious and expensive. Now， researchers take a different approach， using unlabeled data to perform scene flow training， which is relatively easy to generate.

The key to their approach was to develop a way for the system to detect its own errors in scene flow. At each instant， the system tries to predict where each 3D point is going and how fast its moving. In the next instant， it measures the distance between the points predicted location and the actual location of the point near that predicted location. This distance forms one type of error to be minimized. The system then reverses the process， starting with the predicted point location and working backward to map back to where the point originated. At this point， it measures the distance between the predicted position and the actual origination point， and the resulting distance forms the second type of error. The system then works to correct those errors.

The researchers calculated that scene flow accuracy using a training set of data was only 25%. When the data was adjusted with a small amount of real?world labeled data， the accuracy increased to 31%. When they added a large amount of unlabeled data to train the system using their approach， scene flow accuracy jumped to 46%.

[Reading][Check]

1. What does the underlined word “precise” mean in paragraph 1？

A. Attractive. B. Complex. C. Exact. D. Common.

2. What advantage does unlabeled data have over labeled one？

A. It is easy to generate. B. It almost has no errors.

C. It can cover more objects. D. It can gather information quickly.

3. Whats the most important factor about using unlabeled data？

A. Measuring the exact distance.

B. Predicting the speed of 3D points.

C. Checking the errors by the system.

D. Locating the position of objections.

4. Where is the text probably taken from？