抄録
As the aged population is increasing in Japan. The population of young people to support it is decreasing. Consequently, a welfare support system is required that uses robot agents to give physical assistance to handicapped people. In such a system, the person commands an agent directly by gesture, and this instruction is input into the agent via its sensor. The agent recognizes motions from the sensor input, surmises the intention of the motions, and then takes action for the person. If the agent's knowledge is limited to recognizing an instruction and inferring the intention of the motion and the person cannot instruct the agent, then the agent cannot initiate action. For such a welfare support system to be constructed, the agent needs to be able to act automatically, without instruction.
Here, an agent with such a hierarchical knowledge configuration is called a Welfare Intelligent Robot (WIR) . An emotion-oriented interface has been developed for the WIR. Also, an engineering model of knowledge, emotion and intention for the interface has been produced, and applied to a mobile robot (WIR) . A WIR with this interface can act without instruction. Results of experiments have proved the effectiveness of the interface.
