"Method based on rewarding desired behaviors and/or punishing undesired ones. In general, a reinforcement learning agent is able to perceive and interpret its environment, take actions and learn through trial and error" (Carew, 2021)
There are four types of task learning:
Optimisation:
A search for the best action from a set of alternatives based on several criteria. It often uses Genetic Algorithms (GA).
E.g. Land use allocation according to social or environmental criteria; Household travel schedules.
Negotiation:
It has the purpose of agreeing on mutual advantages for different actors. It uses Genetic Algorithm, Batch Normalization and Neural Networks.
E.g. e-commerce negotiations, auctions, supply chain prices and game theory
Prediction:
Attempt to forecast the future, predicting information based on agent simulation that can give outcome values or actions for other agents.
It generally uses Neural Networks.
E.g. Stock market prediction, Hunter - prey dynamics.
Adaptation:
Alternation of agents' behaviour as a result of a change in their Environment.
It usually uses neural networks, Genetic Algorithms and a hybrid process.
E.g. Manufactures adapt production based on the market; Alter the direction or speed of an evacuation simulation.