To operate autonomous driving systems, a large number of various expensive sensors are usually installed on cars. However, Tesla decided to avoid this and build its own autonomous driving system using cameras. This complicates the AI learning process, however it has a number of key benefits and can provide truly complete autonomy for driving, classified as Level 5 autonomy.
Tesla has published a patent 'Estimating object properties using visual image data.' The disclosed invention helps to receive data based on an image captured with a vehicle camera in order to partially identify the distance of an object from a vehicle.
Autonomous driving systems typically rely on mounting numerous sensors including a collection of vision and emitting distance sensors (e.g., radar, lidar, ultrasonic, etc.). By collecting the data captured by each sensor, the system can understand the vehicle's environment and determine how to control the vehicle. However, as the number and types of sensors increase, the complexity and cost of the system increases.
For example, emitting distance sensors such as lidar are often costly to include in a mass market vehicle. Moreover, each additional sensor increases the input bandwidth requirements for the autonomous driving system. Therefore, Tesla began to search for the optimal configuration of sensors on the vehicle. The ideal configuration should limit the total number of sensors without limiting the amount and type of data captured to accurately describe the surrounding environment and safely control the vehicle.
FIG. 5 is a diagram illustrating an example of capturing auxiliary sensor data for training a machine learning network.
The system described in the patent is comprised of one or more processors coupled to memory. One or more processors are configured to receive image data based on an image captured by a vehicle camera. Then the goal is to utilize this data—as a basis of input to a trained machine learning model—to at least in part identify the distance of an object from the vehicle. The machine learning model has been trained using a training image and a correlated output of an emitting distance sensor.
The patent described a training technique for generating highly accurate machine learning results from vision data. Using auxiliary sensor data, such as radar and lidar results, the auxiliary data is associated with objects identified from the vision data to accurately estimate object properties such as distance. In various parts, the collection and association of auxiliary data with vision data is done automatically and requires little, if any, human intervention.
© 2021, Eva Fox. All rights reserved.
We appreciate your readership! Please share your thoughts in the comment section below.