Pose Estimation Methods for Autonomous Vehicles

Abstract

The invention discloses a position and orientation estimating method for a self-driving car. The position and orientation estimating method includes the following steps: firstly, according to the preset sampling period, obtaining the speed of the front wheels of the self-driving car by virtue of a wheel speed sensor, and meanwhile, obtaining the steering wheel angle of the self-driving car by adopting a steering wheel angle sensor; secondly, computing the center steering angle of front axle of the car as per the steering wheel angle of the car; and finally, according to the current speed of the front wheels and center steering angle of the car, conducting iterative computation to obtain the position and course angle of the car at the previous moment, thus obtaining the position and course angle of the car at the current moment. According to the method disclosed by the invention, the position and orientation estimation for the car can be achieved by virtue of the wheel speed sensor and the steering wheel angle sensor, with the help of simple integral algorithm. The method is low in use cost as compared with a method relying on the high precision sensor, the algorithm is simple and easy to realize, and the method can be widely applied to the self-driving industry.

Description

technical field [0001] The invention relates to the field of automatic driving, in particular to a pose estimation method for an automatic driving vehicle. Background technique [0002] The realization of automatic driving technology requires real-time acquisition of vehicle pose information (position and heading angle). At present, most scientific research institutions and OEMs rely on high-precision GPS for positioning, combined with gyroscopes, accelerometers, etc. for automatic driving. For the calculation of the heading angle of the car, some self-driving unmanned vehicles developed by Google rely on the multi-line lidar on the roof for map construction, and use SLAM technology to estimate the pose of the vehicle in real time. These methods require additional expensive sensors. The measurement accuracy is greatly affected by the operating state of the sensor. For example, the accelerometer is susceptible to external interference and the gyroscope is prone to drift after...

AI Technical Summary

Problems solved by technology

[0002] The realization of automatic driving technology requires real-time acquisition of vehicle pose information (position and heading angle). At present, most scientific research institutions and OEMs rely on high-precision GPS for positioning, combined with gyroscopes, accelerometers, etc. for automatic driving. For the calculation of the heading angle of the car, some self-driving unmanned vehicles developed by Google rely on the multi-line lidar on the roof for map construction, and use SLAM technology to estimate the pose of the vehicle in real time. These methods require additional expensive sensors. The measurement accuracy is greatly affected by the operating status of the sensor. For example, the accelerometer is susceptible to external interference and the gyroscope is prone to drift after long-term use. GPS accuracy is affected by tropospheric refraction and communication delay.

Because the existing self-driving car pose estimation methods are all based on high-precision sensors, its high cost makes it difficult to achieve mass production of autonomous driving technology, and at the same time, the accuracy and robustness of the estimated car pose cannot be guaranteed based on a single sensor data. Stickiness, the usual practice is to use multi-sensor data fusion to make up for their respective defects, but the multi-sensor data fusion algorithm is complex, and the pros and cons of the algorithm directly affect the accuracy of the estimated car pose, which has become a technology of automatic driving technology barrier

In general, the current pose estimation method for self-driving cars not only requires the use of high-precision sensors, which leads to high cost, but also the recognition algorithm is more complicated. These two problems greatly restrict the pose estimation of self-driving cars, and also limit the The development of autonomous driving technology

Images