A dynamic path induction method based on multi-objective sarsa learning

Abstract

The present invention proposes a dynamic path induction method based on multi-objective Sarsa learning. The process includes: information initialization; information update; induction path calculation, including normalization of Q vector table, calculation of scalar value based on driver preference, calculation of Boltzmann probability Distribution, the driver chooses the next driving segment according to his personal preference through the roulette method until the driver's vehicle reaches the destination. According to the traffic conditions of the current traffic system, optimize the driving path of the vehicle, improve the efficiency of the traffic system, and alleviate traffic congestion. From a practical point of view, the dynamic path induction of multiple induction targets at the same time is more in line with the induction needs in real life. Considering the driver's induction preference, the driver is provided with a dynamic induction path that meets personal preferences, thereby increasing the acceptance rate of the induction path, further improving the traffic efficiency of the traffic system, and alleviating traffic congestion.

Description

technical field [0001] The invention belongs to the technical field of intelligent transportation, and in particular relates to a dynamic path induction method based on multi-objective Sarsa learning. Background technique [0002] In recent years, with the rapid development of my country's social economy, the number of private cars has continued to rise, and the pressure on urban traffic has increased accordingly. Problems such as urban traffic congestion, congestion, and frequent traffic accidents have also become increasingly serious. In addition, as an important participant in the traffic system, drivers often have multiple induced goals at the same time and have different preferences for different goals during a journey. Whether to consider the driver's personal preference will have a great impact on the acceptance of induced information and thus affect the traffic efficiency of the traffic system. Therefore, from the perspective of alleviating traffic congestion and sa...

AI Technical Summary

Problems solved by technology

However, compared with the single-strategy method, the multi-strategy method will learn a set of optimal solutions to approach the Pareto front every time it interacts with the environment. This process requires a lot of computing time and the corresponding calculation amount is also very large.

And using a multi-strategy method in on-policy learning, the amount of calculation and the large amount of time required for the corresponding solution set make this method unsuitable for dynamic path induction systems

Images