Shulou(Shulou.com)06/03 Report--

Https://www.toutiao.com/a6714594255387492878/

For the autopilot system, if the perception module is compared to human eyes and ears, then the decision planning module is the brain of autopilot.

After receiving all kinds of perceptual information from the sensor, the brain analyzes the current environment, and then issues orders to the underlying control module.

This process is the main task of the decision planning module.

A typical decision planning module can be divided into three levels:

Global path Planning (Route Planning) Behavioral decision level (Behavioral Layer) Motion Planning (Motion Planning) 1, Global path Planning

Global path planning means that after receiving a destination information, combined with high-precision map information and the current pose information of the vehicle, an optimal global path is generated, which can be used as a reference and guide for subsequent local path planning.

The "optimal" here can refer to conditions such as the shortest path, the fastest time, or the need to pass through a specified point. Common global path planning algorithms include Dijkstra and A-Star algorithms, and a variety of improvements based on these two algorithms.

2. Behavior decision-making level

The behavior decision-making layer refers to: after receiving the global path, according to the environmental information obtained from the perception module (other vehicles, pedestrians and other obstacles, traffic signs, traffic lights and other traffic rules), as well as the current driving path and other state information, make specific behavior decisions (such as lane overtaking, following, giving way, parking, entering and leaving stations, etc.)

Many algorithms used in behavior decision layer include: finite state machine, decision tree, rule-based reasoning model and so on.

There are three challenges that need to be addressed at this layer:

The actual traffic scene is ever-changing, how to be universal; the behavior of other traffic participants is uncertain, so we not only need to predict their behavior, but also need to consider the game between this car and other traffic participants. how to deal with the information provided by the perception module can not be 100% accurate and 100% full coverage. 3. Sports planning

Motion planning refers to generating a trajectory that satisfies various constraints (such as safety, ride comfort, dynamic constraints of the vehicle itself, etc.) according to specific behavior decisions, and the trajectory is used as the input of the control module to determine the driving path of the vehicle.

Motion planning can be further divided into local path planning and speed planning, that is, "behavior" is transformed into a more specific driving "track", so that a series of control signals (acceleration, steering wheel angle, gear, lighting, etc.) can be generated to realize the automatic driving of the vehicle.

Local path planning and speed planning algorithms are commonly used in search-based planning algorithms, sampling-based planning algorithms, and direct optimization methods.

Related reading

Introduction of Speed Planning based on Smurt Diagram-Apollo Learning Notes

Introduction to decision-making Technology of Autopilot system-Apollo Learning Notes

Introduction of obstacle avoidance system of unmanned vehicle: obstacle detection, trajectory prediction and obstacle avoidance path generation

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.