Design of Intelligent Control System for Self-Driving Automobile Trajectory Based on BP Neural Network Algorithm

Engineering notes

Key topics: self-driving, control. See the paper for implementation details and experimental results.

Chinese explanation / 中文解读

中文解读待补充：本站会优先为端到端自动驾驶、BEV感知、3D目标检测、轨迹预测、路径规划、LiDAR感知等高价值论文补充中文说明。

Original abstract

SDA (Self-driving automobile) has more precise control ability than traditional human driving, and its control of acceleration, deceleration and steering is more precise, and it has faster response ability than human driving. Because the relationship between input and interference factors on output is quite complex, and it is a high-order nonlinear system, and the uncertainty of internal parameters makes it very difficult to choose a suitable modeling method for vehicle system. In this paper, an optimal design method based on BPNN (BP neural network) is adopted to design the intelligent control system of SDA trajectory. GA (Genetic Algorithm) and batch normalization method are used to improve the neural network (NN). The output of the network acts on the PID controller, and the control parameters of PID are adjusted by NN. The simulation results show that the maximum lateral error is 0.23 m, and the trajectory obtained is basically consistent with the trajectory obtained from the test, which can meet the requirements of vehicle safety and obstacle avoidance. When the speed of the vehicle is accelerated from $10 \mathrm{~m} / \mathrm{s}$ to $30 \mathrm{~m} / \mathrm{s}$, it takes about 5 s for the vehicle to reach the predetermined speed, which meets the error requirements of vehicle speed control. The results show that SDA trajectory intelligent control system based on BPNN algorithm can reduce vehicle sideslip and improve vehicle driving stability.

Links and sources

• Official / arXiv page

Comments