Working on the Emotion Core update it became clear to me that placing responsibility of emotional analysis to Ego-like nodes (a Consciousness, Reflexes, and Instincts) is a wrong approach. It leads to the situation where the developer of the application should specify several types of behavior themself, which is too complicated for development.

I want to implement another approach when concepts themselves contain information about how they should be modified based on a set of the emotion parameters. For example:

1. An Ego-like node sends the concept move with a modifier left:

1
2
3
4

{
    "concept": "move",
    "modifier": ["left"]
}

2. The concept move contains a descriptor with information that: if adrenaline is lower than 5 - add a modifier slower

3. The Concept-to-command interpreter update the concept to:

1
2
3
4

{
    "concept": "move",
    "modifier": ["left", "slower"]
}

4. According the concept descriptors the Concept-to-command interpreter sends commands to the device Moving Platform:

• Set speed to 2 (3 is maximum for the device)
• Move left
• Set speed to 3 (default value)

In addition, I updated the diagram itself by structuring it and adding notes to make the entire system easier to understand. Here is the diagram:

Now I will implement the structure above in the emotion_core branch of the zakharos_core repository. The feature is getting closer to implementation. More updates soon.

Links:

https://github.com/an-dr/zakharos_core/tree/feature/emotion_core

https://github.com/an-dr/r_giskard_EmotionCore