Needs, Pains, and Motivations in Autonomous Agents

Area of Science:

• Artificial Intelligence
• Machine Learning
• Agent-Based Systems

Background:

• Previous research established a motivated learning (ML) agent with symbolic input/output.
• Enhancements focused on granting the agent autonomy for inter-agent interaction.

Purpose of the Study:

• To explore the autonomous development of motivations and memory in artificial agents.
• To investigate agent-based learning within a simulated environment.

Main Methods:

• Development of a motivated learning (ML) agent with symbolic input/output.
• Implementation of drives and pains to establish agent motivations and abstract goals.
• Simulation of the ML agent within the NeoAxis game engine.
• Comparative analysis against various reinforcement learning (RL) algorithms in a dynamic test scenario.

Main Results:

• The ML agent exhibited autonomous development of motivations and memory.
• The agent successfully learned to respond to desired and undesired events, creating abstract goals.
• The ML agent outperformed all tested reinforcement learning (RL) algorithms in the dynamic test scenario.

Conclusions:

• The developed ML agent demonstrates enhanced learning capabilities through autonomous motivation and memory development.
• The findings suggest that ML agents offer a promising alternative to traditional RL algorithms in complex environments.
• This work contributes to the advancement of autonomous intelligent agents capable of sophisticated interaction and learning.