Path navigation method and system based on reinforcement learning

Abstract

The invention discloses a path navigation method and system based on reinforcement learning. The method comprises the following steps: constructing a road adjacency relationship diagram of a city according to city map data; predicting congestion indexes of different road sections of different periods of time of the city according to vehicle trajectory data and the road adjacency relationship diagram; constructing a road congestion probability diagram of the city according to the congestion indexes based on the road adjacency relationship diagram; and generating a navigation path based on the reinforcement learning, wherein the state space of the reinforcement learning comprises the road congestion probability diagram. The method makes the urban road congestion probabilistic, more intuitiveand easy to visualize on the basis of numerical value; the road congestion calculation only utilizes the road condition and the historical vehicle trajectory data to enable convenient practice; differing from the common obstructed path finding method, the probability road finding value is more accurate, and the route that the common path finding algorithm cannot find is discovered; the reinforcement learning is used as a heuristic algorithm to consider the time-consuming and unobstructed path finding, so as to obtain the global optimal solution and increase the accuracy of the path finding algorithm.

Description

technical field [0001] The invention belongs to the technical field of path navigation, and more specifically, relates to a path navigation method and system based on reinforcement learning. Background technique [0002] Mobile phone navigation to find effective driving routes has become a daily routine. A good driving route not only saves the driver's time, but also saves energy consumption. The widespread use of GPS devices allows us to easily obtain city road details such as traffic flow, speed, etc. These data play an extremely important guiding role in path navigation. [0003] In the prior art, the patent CN108847037A discloses a non-global information-oriented urban road network path planning method, which enables the road network to have the ability of adaptive adjustment to the distribution of traffic flow through reinforcement learning, so the state of the road network is in a flow balance state . However, the A*R pathfinding algorithm in this method uses a rou...

AI Technical Summary

Problems solved by technology

[0005] Aiming at the defects of the prior art, the purpose of the present invention is to optimize the prior art in the pathfinding algorithm with many preconditions and the problem that the decision function is not perfect enough

Images