Investigating Case Learning Techniques for Agents to Play the Card Game of Truco

Abstract

Truco is a popular game in many regions of South America; however, unlike worldwide games, Truco still requires a competitive Artificial Intelligence. Due to the limited availability of Truco data and the stochastic and imperfect information characteristics of the game, creating competitive models for a card game like Truco is a challenging task. To approach this problem, this work investigates the generation of concrete Truco problem-solving experiences through alternative techniques of automatic case generation and active learning, aiming to learn with the retention of cases in case bases. From this, these case bases guide the playing actions of the implemented Truco bots permitting to assess the capabilities of each bot, all implemented with Case-Based Reasoning (CBR) techniques.