Task classification in multi-role robotic systems for independent living

Homecare robots aiming at supporting the elderly independently living are necessary for aging society. Such kind robots must have multi-role with various capability to provide essential services, such as laundry, setting up table and preparing dinner. Each of these services is a task to be fulfilled from the view point of homecare robots. Understanding each task and being able to differentiate tasks are crucial for a homecare robot to successfully provide desired assistance and service to aging population. This paper introduces a Nature Language based approach to task classification. A verb-based task classifier was developed to extract tasks in the form of verbs and nouns given user´s commands. A verb vector technique was then employed to classify tasks into categories according to verb clusters. Experiment results demonstrated that the proposed task classifier can successfully extract tasks from user´s commands and to classify them into task categories.