MIND T3_player agent: A Self-Taught Agent in a Tic-Tac-Toe environment
The T3-Player is a self-taught agent who inhabits a dynamic tic tac toe environment where a differential-difference (time-lag) process has to be solved. Each process is an episode (match) during which two agents O an X randomly play against each other by alternating turn. There exist auxiliary, background routines belonging to the environment that checks for winning positions and return reward information to the environment.
T3-Player is an agent from the root and satisfies the conditions:
- Perceives its environments through sensors (board state neurons).
- Acts upon that environment through effectors (token insertion neurons).
- Act autonomously (i-agent neurons).
- Realize a set of goals or tasks to learn to play top-level games. (Gradient Descend and Reinforcement Learning).
The current T3-Player model is constructed and delivery in windows Borland C-C++, this is an easy to install, friendly ambient that produces stable, stand-alone .exe code. Linux C version can be directly constructed using our C code and converting the graphic routines from Borland C to Linux C (contributions are welcome).
As a noticeable advantage, our C code runs almost directly in ARDUINO boards. (running code coming soon).
The overall T3-Player agent system for Tic-Tac-Toe is delivered in three modules. You can download each module in the following sections:
The neural networks and gradient descend come into action.
Bellman equation and the self-taught agents appears, creating human level artificial players.
Multi joint robot driven by a self-taught agent
In previous research we proposed and develop computer models written in C++, based on Artificial Intelligence, for a self-motivated, multi-joint virtual robot driven by an agent that learns by itself to control the robot’s body and execute complex folding motions that produce intelligent mass displacements in multifactorial environments .
The robot’s mechanical joints, muscles and sensors are controlled by trainable artificial neurons and its abilities to execute challenging mechanical work are acquired by a self-taught agent, a computer program capable of exploring and learn new solutions by itself through a combination of:
- Artificial Neural Net.
- Gradient Descent.
- Reinforcement Learning. Algorithms.