Behavior Selection Algorithm for Personal Service Robots Using Intelligence Operating Architecture

Personal service robots assisting humans in daily life are expected to provide a user-oriented service [1]. To meet this expectation, there has been much research on behavior selection algorithms for the proper service. The action selection method was proposed to activate a most appropriate action in a certain situation [2]. The behavior selection method for artificial creatures considering both internal and external situation was proposed [3]. A system using episodic memory (EM) was developed to supervise user’s daily activities [4]. However, the agents in the existing research did not much consider user’s past requests or behavior patterns in making a decision for a proper behavior. In this research, the intelligence operating architecture (iOA) is used to construct a user oriented behavior selection algorithm [5]. The iOA consists of five parts and 15 modules to implement the robot intelligence. It is motivated by the key functions of the human brain, such as problem solving in the frontal lobe, actions from the motor cortex, etc. Among them, context generation, EM, semantic memory (SM), decision making modules are the key modules to provide users with autonomous user-oriented services. The context generation module identifies the current situation by “when,” “where,” “what,” etc. The EM stores the events along with time and the SM holds the factual and concept-based knowledge. By referring to EM and SM, the most proper behavior of a robot in a given situation is selected in the decision making module. The detailed procedure of the proposed algorithm is described in the following. To begin with, the context module defines the current event by five attributes, i.e. ‘user ID,’ ‘behavior,’ ‘time,’ ‘place’ and ‘object.’ If the current event is a user’s request and the given information is not sufficient, the robot searches for the most similar event in the EM. The similarity degree Sim(ei, ej) between the two events, ei and ej , is calculated as follows: