On-line estimators for ad-hoc task execution: learning types and parameters of teammates for effective teamwork

Abstract It is essential for agents to work together with others accomplish common objectives, without pre-programmed coordination rules or previous knowledge of the current teammates, a challenge known as ad-hoc teamwork. In these systems, an agent estimates algorithm in on-line manner order decide its own actions effective A approach assume set possible types and parameters reducing problem into estimating calculating distributions over types. Meanwhile, often must coordinate decentralised fashion complete tasks that are displaced environment (e.g., foraging, de-mining, rescue fire control), where each member autonomously chooses which task perform. By harnessing this knowledge, better estimation techniques can be developed. Hence, we present On-line Estimators Ad-hoc Task Execution (OEATE), novel teammates’ type parameter execution. We show theoretically our converge perfect estimations, under some assumptions, number increases. Additionally, run experiments diverse configuration level-based foraging domain full partial observability, “capture prey” game. obtain lower error than approaches performance completed cases. fact, evaluate variety scenarios via increasing agents, scenario sizes, items, types, showing overcome works most cases considering process, besides robustness even erroneous potential