Flocking Algorithm And Its Application In Cooperative Autonomous Driving

Flocking phenomenon is a ubiquitous but very complex phenomenon in nature, and the corresponding flocking algorithm plays a great role in promoting the development of distributed control theory. Many research products are widely used in the field of robots formation, UAVs(Unmanned Aerial Vehicles), mobile sensor networks etc. Flocking algorithm used in cooperative autonomous driving has also been a concern of researchers in recent years. This thesis mainly discusses flocking algorithm and its application in cooperative autonomous driving. The contents can be divided into three parts:1. An improved flocking algorithm with a minority of informed agents. The mutual influence between agents was achieved by the potential function, but the potential function could not be infinite in practice, which might lead to broken links among some agents and could not achieve the ideal flock. To solve this problem, this paper introduced navigation items to force the agents tend to the center of mass, and based on this proposed an improved flocking algorithm with a minority of informed agents. Through theoretical analysis, the stability of the algorithm and velocity matching were proved. To verify its practicability, the improved algorithm is applied to mobile sensor network. Through theoretical analysis and simulation, it proved a minority of sensors was activated in networks can make all the sensors tracking target, and this will greatly reduce the power consumption of the mobile sensor network.2. Flocking algorithm with a minority of informed agents based on limited field of view. This thesis used sensor model with limited field of view instead of the omnidirectional detection model that used in the traditional multi-agent system, the sensor is rotated in a constant angular velocity, and the potential energy function is still used to achieve interaction between agents during the rotation. Through stability analysis and simulation, flocking algorithm based on limited field of view can be achieve Lyapunov stability but the speed of agents will concussion in the speed of the leader in a small amplitude according to the rotation cycle of sensor. Followed, the sensor model with limited field of view is applied to the flocking algorithms with a minority of informed agents that presented in the previous section. We analyzed the sensor model with limited field of view is still able to accommodate a minority of informed agents.3. Cooperative autonomous driving algorithm based on limited field of view. Considering the sensor model with the limited field of view in vehicle, and use it in cooperative autonomous driving algorithm that based on traditional flocking algorithm. The lane-driving model, lane-changing model and braking model in algorithm were achieved in simulation. Finally, we used the cooperative autonomous driving algorithm achieved two of classical traffic scene by nonlinear dynamics switch method, verified the effectiveness of the algorithm.