An objective of the present application is to provide a task optimization method and device in a mobile robot. The method in the present application comprises: obtaining at least one task type in a mobile robot and use information when all users use tasks corresponding to each task type; respectively performing machine learning on the use information of all the users corresponding to each task type, so as to obtain at least one piece of user use habit information corresponding to each task type and a use probability thereof, and performing machine learning on the use information when all the users use the tasks corresponding to the task types; and based on the at least one piece of user use habit information corresponding to each task type and the use probability thereof as well as real-time use information, optimizing the tasks that correspond to the task types and that are used by the users in real time. In this way, a mobile robot can be helped to more efficiently and intelligently complete execution of the tasks corresponding to the task types, and time and resources spent when the users use the tasks corresponding to the task types in the mobile robot in real time can be reduced.