An intelligent vehicle GPU parallel acceleration trajectory planning method

Abstract

The invention discloses an intelligent vehicle GPU parallel acceleration trajectory planning method, and the method comprises the steps of obtaining a real-time perception obstacle grid graph, a global reference path, intelligent vehicle real-time GPS information and an upper layer decision instruction at a CPU end, carrying out the sampling along the reference path, and obtaining a series of target sampling terminals; at the GPU end, obtaining a sampling terminal sequence, the real-time state information of an intelligent vehicle, the perception environment grids and the historical frame planning track parameters, designing a track generation kernel function, generating an intelligent vehicle stable tracking track connected with an initial state and a terminal state for each target sampling terminal, designing a track evaluation kernel function, and evaluating the cost of each track; and returning the track data to the CPU end, selecting an optimal track and matching the optimal trackwith the speed value, obtaining a track planning result and transmitting the track planning result to the control execution mechanism. The method is used for improving the real-time planning efficiency of the track of the intelligent vehicle, achieving the purposes of efficiently generating a large number of tracks in real time and further improving the track planning result.

Description

technical field [0001] The present invention relates to the technical field of intelligent transportation systems, in particular to a trajectory planning method for GPU parallel acceleration of an intelligent vehicle, in particular to a real-time trajectory planning method in an unmanned driving system, specifically to realize the parallel design of trajectory generation and cost evaluation in the cuda architecture , which is used to improve the efficiency of trajectory planning, and plan as many trajectories as possible under the premise of satisfying the real-time performance of intelligent vehicles to improve the final planning results. Background technique [0002] In recent years, a large amount of work has been devoted to solving the problem of motion planning for unmanned vehicles. These methods can be broadly classified into two categories: graph search based methods and trajectory generation based methods. Graph search-based methods focus on computing collision-fre...

AI Technical Summary

Problems solved by technology

Because it solves the constrained optimization problem in the continuous control space at the level of long-term prediction, it usually involves complex optimization process, including matrix operation and numerical iterative operation, which easily leads to excessive computational burden, and the irregular obstacle distribution in the real-time environment is caused by The imposed constraints are also difficult to handle, making it difficult to achieve real-time trajectory planning needs

Images