A Review of Autonomous Vehicle Path Tracking Algorithm Research
Yaoting Chen, Yanping Zheng *
College of Automobile and Traffic Engineering, Nanjing Forestry University, Nanjing 210037, China;
carmelo@njfu.edu.cn (Y.C.); qq1137656476@163.com(Z.P.);
* Correspondence: zhengyp@njfu.com.cn; Tel.: +86-138-5186-4173
Abstract : Driverless technology aims to improve driving safety, accuracy and comfort. Path tracking is a basic component of the motion control module of autonomous vehicles, and its control algorithm directly affects the path tracking effect. Based on the preliminary results of the application of path tracking control algorithm, this paper analyzes the principles, advantages and disadvantages, applications and current research progress of the path tracking algorithm under different working conditions from the perspective of different working conditions at low speed and high speed, and provides an outlook on the future development, aiming to provide reference for future in-depth research.Keywords : Autonomous vehicles; Motion control; Path tracking; Pateral control
Introduction
In recent years, autonomous vehicle have become increasingly popular in the automotive markets, which mainly consist of task decision, environment sensing, path planning, path tracking and vehicle control subsystems [1,2]. As a crucial part of the architecture of driverless vehicles, path tracking control is undoubtedly a key research focus for scholars from various countries. The role of path tracking control is to control the actuator action based on the vehicle state information given by each on-board sensor and the planned reference path to ensure that the vehicle travels along the planned path, and to control the actuator in time to reduce the deviation between the vehicle position and the planned path when the vehicle deviates from the planned path [3]. Therefore, the accuracy of path tracking control directly determines whether the driverless vehicle can follow the planned trajectory, which is of great significance for traffic safety. Many research results have been generated around path tracking algorithms in recent years, and there are common problems and solutions in solving the problem of reducing tracking errors.
In this paper, the path tracking control algorithms are divided into two categories for low-speed and high-speed conditions according to the speed of the car. Low-speed conditions mainly refer to the driving environment where the vehicle speed is lower than 30km/h, including scenarios such as factories, ports, campuses, farms, mines, and parking lots [4,5]. Low-speed driverless cars do not need to consider the influence of vehicle dynamics characteristics due to low vehicle speed, and only need to control the accuracy of path tracking, which requires high control accuracy [6-8]. High-speed conditions mainly refer to the driving environment where the vehicle speed is higher than 30 km/h. Due to the high speed and the existence of high-speed lane changing and high-speed cornering conditions, the influence of vehicle dynamics on the maneuvering stability needs to be considered when path tracking control is carried out, and the control effect directly affects the driving safety of the vehicle [9].
Although the same path tracking control algorithm can be used for driverless vehicles under both low-speed and high-speed conditions, the path tracking control faces different challenges and the controller design differs due to the difference in control objectives. Under low-speed conditions, considering the motion characteristics of the vehicle, the path curvature of the vehicle driving is larger and the vehicle heading angle changes more, which makes the optimal solution of the nonlinear problem more difficult and leads to the problem of low path tracking accuracy [10]. Under high-speed conditions, due to the contradiction between the complexity of the vehicle dynamics model and the computational real-time, driverless vehicles in high-speed situations need a longer prediction time domain for hazard avoidance, further increasing the difficulty of vehicle controller design [11,12]. Therefore, in this paper, the problems faced by path tracking control and their algorithms are sorted out according to two operating conditions, low-speed and high-speed, respectively, and the characteristics of different algorithms are analyzed with the aim of providing references for future in-depth research.
2. Research status of low-speed path tracking algorithm
The kinematic characteristics of the vehicle are more influential when the driverless vehicle is at low speed, and due to its own small lateral acceleration and the existence of the minimum turning radius constraint, the influence of vehicle dynamics is generally not considered, i.e., there is no need to consider the maneuvering stability of the vehicle, and the kinematic model of the vehicle can be directly used as the control model of the path tracking algorithm, at which time the path tracking controller designed based on the kinematic model has reliable control performance [ 13-16]. At present, the path tracking algorithms that are more applied in low-speed conditions are PID control method (Process Identifier Derivative, PID), pure tracking control algorithm (Pure Pursuit, PP), Stanley algorithm, model predictive control (Model Predictive Control, MPC), etc. [ 17-19].
2.1 PID algorithm
The PID control method is a widely used automatic controller that controls by proportional (P), integral (I), and differential (D) of deviation, which has the advantages of simple principle, easy implementation, wide applicability, and relatively independent control parameters [20,21]. Figure 1 and Figure 2 show the PID control schematic and feedforward PID control framework, respectively, in which the proportional link can reflect the state deviation of the controlled object in a timely and proportional manner, the integral link improves the non-differential degree of the controlled object through the memory of the state deviation, the differential link can reflect the trend of the state deviation to speed up the system action, and the feedforward control can compensate the disturbance and weaken the effect of the disturbance on the system output [22]. of the system [22].
Currently, many studies are based on PID algorithms or joint control with PID for path tracking. For example, the literature [23,24] designed a path tracking controller based on PID algorithm and verified the advantages of this algorithm in low speed conditions. Based on the idea of hierarchical control, the literature [25] designed the method of joint control of PID and MPC and verified that the controller has good tracking effect in low-speed conditions. Considering the characteristics of low speed and small curvature working conditions, literature [26] designed a transverse controller based on neural network parallel PID, and used PID algorithm to adjust the error caused by the neural network training to further improve the control accuracy.
For the problem of difficult adjustment of PID controller parameters, literature [27] proposed a method based on the combination of PID control and pre-scanning control, and designed a PID controller with adaptive parameter values using fuzzy inference control.Mashadi B et al. introduced a genetic algorithm to adjust the parameters and compared the control effect of the traditional PID controller, proving that the controller has better tracking accuracy, real-time performance and vehicle driving stability [28].
PID control is simple and effective, but due to the delay between the sensor and the actuator, the PID controller can only get the position deviation of the previous moment, so there is an unavoidable delay in the system, and this error is not negligible when the vehicle speed is fast, so the PID algorithm is poorly adaptable and vulnerable to the change of the road environment, and each parameter is difficult to adjust and cannot meet the requirements of path tracking under high-speed conditions, so In practical applications, it is generally used in conjunction with other control algorithms.