Most intelligent transportation systems, such as driverless cars and drones, try to reduce people’s participation and realize intelligent transportation. However, the lack of human participation will lead to the weak adaptive ability of intelligent transportation systems, which can't cope with emergencies and other problems. In response to this situation, this paper introduces the adaptive decision-making method of human-in-the-loop. Firstly, the agent is trained by introducing human feedback into the traditional reinforcement learning algorithm. Through the interaction between agents with the environment and humans, the process of "exploration-learning-decision making" is repeated constantly, accumulating experience and optimizing strategies in the process of interacting with the environment and humans. The agent continuously updates the decision-making process through human feedback and can avoid static obstacles and dynamic obstacles in the environment, finally reach the target point. Secondly, the human-in-the-loop algorithm of DQN-TAMER is put forward, and experiments are carried out through three groups of human-in-the-loop algorithms. The experimental results show that the self-adaptive decision-making method of human-in-the-loop traffic can obviously improve the decision-making ability of intelligent transportation systems and the adaptability of the whole system, and the learning efficiency of agents is significantly improved by DQN-TAMER's human-in-the-loop algorithm. Finally, the medical material transportation simulation system is used as a prototype.
Research on Human-in-the-loop Traffic Adaptive Decision Making Method
2022-09-25
4720001 byte
Conference paper
Electronic Resource
English
Adaptive channels and human decision making
Tema Archive | 1984
|Human Factors In Air Traffic Controller Decision Making
British Library Conference Proceedings | 1990
|