SIGNAL CONTROL APPARATUS AND SIGNAL CONTROL METHOD BASED ON REINFORCEMENT LEARNING

A signal control apparatus and a signal control method based on reinforcement learning are presented. The signal control apparatus may comprise: a capturing unit for acquiring an intersection image by capturing an image of an intersection; a control unit for calculating information about a congestio...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
1. Verfasser: CHOI, Tae Wook
Format: Patent
Sprache:eng ; fre ; kor
Schlagworte:
Online-Zugang:Volltext bestellen
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:A signal control apparatus and a signal control method based on reinforcement learning are presented. The signal control apparatus may comprise: a capturing unit for acquiring an intersection image by capturing an image of an intersection; a control unit for calculating information about a congestion level at the intersection by analyzing the intersection image acquired by the capturing unit and calculating control information regarding a next signal by using the information about the congestion level and a reinforcement learning model trained by using, as input values, state information and reward information calculated on the basis of the congestion level at the intersection; and a driving unit for driving a traffic light according to the control information regarding the next signal, calculated by the control unit. La présente invention concerne un appareil de commande de signal et un procédé de commande de signal basés sur l'apprentissage par renforcement. L'appareil de commande de signal peut comprendre : une unité de capture pour acquérir une image d'intersection par capture d'une image d'une intersection ; une unité de commande pour calculer des informations concernant un niveau de congestion au niveau de l'intersection par analyse de l'image d'intersection acquise par l'unité de capture et pour calculer des informations de commande concernant un signal suivant en utilisant les informations concernant le niveau de congestion et un modèle d'apprentissage par renforcement entraîné en utilisant, en tant que valeurs d'entrée, des informations d'état et des informations de récompense calculées sur la base du niveau de congestion au niveau de l'intersection ; et une unité d'entraînement pour entraîner un feu de circulation en fonction des informations de commande concernant le signal suivant, calculées par l'unité de commande. 강화학습 기반 신호 제어 장치 및 신호 제어 방법을 제시하며, 신호 제어 장치는 교차로의 영상을 촬영하여 교차로 이미지를 획득하는 촬영부; 상기 촬영부가 획득한 교차로 이미지를 분석하여 상기 교차로의 혼잡도 정보를 산출하고, 교차로의 혼잡도에 기초하여 연산되는 상태 정보와 리워드 정보를 입력 값으로 하여 트레이닝 되는 강화학습 모델과 상기 혼잡도 정보를 이용하여 다음 신호에 대한 제어 정보를 산출하는 제어부; 그리고 상기 제어부가 산출한 다음 신호에 대한 제어 정보에 따라 신호등을 구동시키는 구동부를 포함할 수 있다.