Kanda, T. et al. (2004). Development and Evaluation of Interactive Humanoid Robots.

Summary

This paper presents our exploratory approach for the development and evaluation of interactive humanoid robots. Our evaluation approach is to measure the body movement interaction between a humanoid robot and humans and compare the results with traditional subjective evaluation.

The robot is 120 cm to reduce intimidation on humans. And has movement in the arms ( 4*2 degrees of freedom (DOF)), the head (3 DOF), the eyes (4 DOF). The robot has various sensors: 16 skin sensors covering the major parts of the body, 10 tactile sensors around the mobile platform, an omnidirectional vision sensor, two microphones to listen to human voices and 24 ultrasonic sensors for detecting obstacles.

We think there is a strong correlation between the numbers of appropriate behaviors an interactive robot can produce and its perceived intelligence. Our constructive approach (bottom-up) is to continue to implement behaviors until humans think that the robot has an animated and lifelike existence beyond that of a simple automatic machine. This is a bottom—up approach, as opposed to a top-down approach, in which human behavior is first analyzed and projected onto a robot.

The behavior of a robot is predetermined and is handled by utilizing modules that modify and transition the internal state of the robot. If a robot has a conversation with a person the robot’s behavior is ‘selected’ by the situated modules. If suddenly a bell rings or another sudden impulse is received from the environment the robot’s internal state transitions to a reactive module to react to the sound of the bell. Input from sensors are preprocessed at sensor modules such as speech recognition. Actuator modules perform low-level control of actuators.

Tests to evaluate the developed robot and analyzed the interaction between the robot and the humans were made. During these tests the humans behave as if they were interacting with another human rather than a robot. The humans kept eye contact with the robot and imitated gestures of the robot among other forms of interaction. The level of interaction that is possible with the robot indicates a high performance of the developed robot.