27/11/2023
Let's say we want to design an Autonomous Driving system. This means that we want the car to be able to drive autonomously from point A to point B while adapting to the possible unpredictable events or obstacles that may occur during the trip. As low-level control, we can only accelerate, brake, and steer the car. As a way to capture information, we have sensors such as cameras, LiDAR (Light Detection and Ranging) detectors, radars, and GPS. Typically, the problem can be broken down into the following pieces:
- High-level path planning: based on the current location and state of the car within the environment, we need to establish a series of small steps to take us from A to B. This means establishing all the different roads and intersections that we need to go through, as well as the different lane changes that will need to happen to turn at the different intersections safely.
- Behavior arbitration: at any point of the trip, the system needs to be able to adjust in real-time to unplanned events. You can think of behavior arbitration as small real-time corrections to the pre-planned path. This module makes decisions on actions to take to bring the passengers to their destination safely. The different actions could be Line-changes, U-turns, Emergency stops, Merging onto highways, Overtaking…
- Low-level planning: once the action has been planned, either by the High-level path planning module or the Behavior arbitration module, the optimal trajectory has to be calculated by taking into consideration the current speed and location of the car and the surrounding vehicles and obstacles.
- Motion controllers: Once the trajectory has been calculated, it has to be executed using the different controls available. The motion controller translates the calculated trajectory into a set of low-level controls such as acceleration, braking, and steering.
- Localization: to accurately establish decisions, the system needs to know. For Mr. Gelson Jacinto
