Project Terminology

Term(s) Meaning
Virtual Environment The playground in which a reality is simulated.
Situational awareness Having contextual knowledge of the current situation
Perceptual understanding Understanding via sense/perception to obtain understanding.
Simulated artificial entity Agent, Game Character, NPC, Robot, an abstraction of a human intelligence.
Agent reasoning repertoire Cognitive capabilities that the agent can use.
Agent Simulated abstraction of intelligence.
Knowledge Acquisition The ability to use information to make logical choices/decisions and inform reactions.
Reasoning Interpreting a situation and being able to choose/use/select knowledge that applies to it.
Situational Rules Contextual and situational Patterns, Repeatable cause & effect
Situational Circumstance Information that defines circumstance(s) which make up a situation.
Experiential An active process or experience of looking for, or obtaining information about, varying contextual situations.
Perceptual, experiential knowledge Knowledge that is obtained through sources of perception.
Adaptable, perceptive, reactive Adaptable: behaviour and understanding might change based on different knowledge available. Perceptive: Utilising continual perceptual monitoring. Reactive: Showing a response to a stimulus.
Real-time using a continual feedback processing model.
Actionable situational knowledge The ability to use that knowledge in a detected situation
Instinctive Based on rules for behaviour in circumstances.
Emergent Rules Rules can be derived, tuned, and evolve according to an ongoing process of rule detection/creation.

Online Learning

Term Meaning
Markov Chain (MC) A model of state transitions represented as a Transition Matrix
Markov Decision Process (MDP) A model a a game as described in Markov Decision Processes
Q-Learning An approach to reinforcement learning that aims to solve MDPs
Q-Value A value metric for an action/move
Policy A decision-making rule or action selection strategy
Decision A choice that selects an Action
Deterministic Policy Selects an action
Stochastic Policy Provides possible actions as probabiliites ie Left 0.7, Right 0.3
Online-learning Dynamic dataset
Offline-learning (aka batch learning) Static/Fixed dataset
Model free Does not learn or use the rules of the envionment ie environment transition dynamics
Model-based Uses or learns the rules of the environment (MDP's Transition Probabilities or MC's Transition Matrix)

 

General Ideas

Term

 Meaning

Experience

 Actions and results (cause and effect)

Use of historical data

 Knowledge

Control policies

 Decision-making strategies
Minimizing failed expectations Need for learning
Goals/Rewards Often tied to Policies that aim to reach goals

Psychology

 Understanding behaviour and what causes it

Minimizing failed expectations

 Need for learning

Goals/Rewards

 Often tied to Policies that aim to reach goals