version 0.272

Adherent to pyMDQN module

Reinforcement Learning atributes:

• 4 agent actions:

  • wait;
  • look;
  • handwave;
  • handshake.

• 5 reward types:

  • fail handshake: -0.1 (
  • successful handshake: 1 (adjustable);
  • fail wave (human ignored): -0.1 (adjustable);
  • successful eyegaze (human makes eye contact) : 0.2 (adjustable);
  • neutral: 0.

• States:

  • 8 sequential grayscale images;
  • 8 sequential depth images.