Artificial Creatures "Rity" and "Humanoid"

attachment:rity_fig10.jpg

RITY, an RIT lab's artificial creature, has its own motivation, homeostasis and emotion, and decides its behavior from these. Through the interaction with a user, it can have its own character. An emotion, expressed mathematically by psychological studies, is changed by a previous emotional value and external information. An emotion is formed by "Happiness", "Sadness", "Fear", "Angry". RITY shows its emotion with a facial expression. A various behavior is calculated by a motivation and homeostasis. RITY carries out a proper behavior to satisfy its motivation and homeostasis.

attachment:rity_fig11.jpg

RITY can learn what to do, when a user orders it, by an interactive learning methods. RITY also can learn what feels good or what feels bad from a previous experience. Training RITY is similar to train a pet. According to a user's command, RITY gets a rewarding or a penalty after its behavior. RITY is learned by this repeated process.

attachment:dd-explain.jpg

attachment:dd-explain2.jpg

  • Media Player will be located here.

Behavior Selection Method for Artificial Creatures Using Degree of Consideration-based Mechanism of Thought (DoC-MoT)

To have artificial creatures interact with their environment like living creatures, a proper behavior selection method is needed. For this purpose, the behavior selection method for artificial creatures mimicking the mechanism of thought of human beings, i.e. degree of consideration-based mechanism of thought (DoC-MoT), is proposed. Internal wills and external contexts were defined as input symbols as assumed facts, and behaviors were assigned as target symbols. The values of knowledge links between input and target symbols were represented by partial evaluation values of target symbols over each input symbol. The DoCs for input symbols were described by fuzzy measure, and the global evaluation of the target symbols were achieved by fuzzy integral aggregating the DoCs and partial evaluations in the memory. To demonstrate the effectiveness of the proposed method, experiments were carried out for an artificial creature, “Rity.” The reasonable behaviors considering the will and context were selected by the proposed method, and Rity was successfully trained to do proper behaviors and to have desired characteristics through the learning process.

Demo 1: Changing Characteristics of Rity by Punishment (Download)

Media Player will be located here.

Demo 2: Changing Characteristics of Rity by Reward (Download)

Media Player will be located here.

Artificial Creatures in Virtual Environment

They are generated by Evolutionary Generative Process for a Personality (EGPP), where blue Rity is an agreeable one and brown Rity is an antagonistic one. The agreeable personality assumes strength in curiosity, intimacy and happiness, and weakness in greed, desire to control, avoidance, anger and fear. In contrast, the antagonistic personality assumes weakness in curiosity, intimacy and happiness, and strength in greed, desire to control, avoidance, anger and fear.

In a video clip, agreeable character shows active behaviors and happy expressions. However, you can observe that antagonistic character shows careful and surprised actions such as stepping backward and angry expressions at times.

When “pats” are given to two Ritys by imposing single clicking, agreeable Rity shows happy expression immediately rather antagonistic Rity changes to happy face after receiving many times of pats.

When “hits” are given to two Ritys by imposing double clicking, you can observe that the emotion of agreeable and antagonistic character change to sad and angry, respectively.

When "dandlings" are given to two Ritys by dragging Rity slowly and softly at the last part of clip, their faces change to happy expression. But the antagonistic Rity changes to neutral expression before long.

[Demo video file]

Media Player will be located here.


CategoryResearch

KAIST logo

373-1, Gusung-Dong, Yusung-Gu, Daejeon. 305-701 Republic of Korea

Tel: +82-42-350-5448(Room #3239), 8048(Room #3238), 3448(Professor)

Emergency call: 010-4395-2210

COPYRIGHT©-2008 Robot Intelligence Technology Lab. / Department of Electrical Engineering