Robot trajectory prediction method based on model confidence and Gaussian process

Abstract

The invention relates to a robot trajectory prediction method based on the model confidence and Gaussian process, and belongs to the field of target tracking and artificial intelligence. The method aims at solving the problems that an existing learning-based method seriously depends on a training data set, and trajectory prediction precision in a data sparse area is low. The method includes the steps that firstly, a robot decision probability model based on the model confidence is established; then, an augmented data set is obtained through a probability path generated by the model; and a Gaussian process model is trained by using the augmented data set, and a robot long-term motion trajectory prediction algorithm is designed in combination with the Gaussian process and RRT, so that the robot long-term motion trajectory prediction precision under the condition of data sparsity is improved.

Description

technical field [0001] The invention belongs to the field of target tracking and artificial intelligence, and in particular relates to a robot trajectory prediction method based on model confidence and Gaussian process. Background technique [0002] With the development of science and technology, robots are widely used in industrial manufacturing and people's lives. Robots are an important entry point for improving human life. Vigorously developing the robot industry is of great significance for promoting industrial transformation and upgrading and improving people's living standards. Robots first need to consider obstacle avoidance in the process of performing tasks. Path planning is an important link for robots to work safely and efficiently. Trajectory prediction, as an important part of the threat assessment system, is one of the key research issues in the field of robot planning. Trajectory prediction can be applied to the anti-collision detection system of unmanned veh...

AI Technical Summary

Problems solved by technology

[0005] In order to solve the problem that the existing learning-based method relies heavily on the training data set and the trajectory prediction accuracy is low in the data sparse area, the present invention provides a robot trajectory prediction method based on model confidence and Gaussian process. The robot decision-making probability model of model confidence, and then use the probability path generated by the model to obtain the augmented data set, and use the augmented data set to train the Gaussian process model, and combine the Gaussian process and RRT to design the long-term motion trajectory prediction algorithm of the robot, which improves the data sparseness. The long-term motion trajectory prediction accuracy of the robot in the case of

Images