The invention relates to the technical field of intelligent
robot behavior control, and discloses a cleaning
robot task self-learning control method, which comprises the following steps that a user sends a task instruction to a cleaning
robot, the robot generates a task process after analyzing the instruction, and the robot acquires
working environment data in real time and performs
task learning on the cleaning robot. A cleaning task model is established through a deep
reinforcement learning algorithm DQN, cleaning
control parameters are generated, cleaning work is carried out, task execution data are collected, the cleaning
control parameters are dynamically adjusted according to the task execution data in the working process, after the cleaning work is completed, a user can submit
user feedback information, and user experience is improved. And finally, task execution data and
user feedback are uploaded to a
cloud server, and a user personalized model is established. Through multi-
modal data fusion, the behavior evaluation model and
visual interaction, the dynamic adjustment and self-learning ability of the cleaning robot in a complex environment is realized, and the cleaning effect, the cleaning efficiency and the user experience are remarkably improved.