RT Dissertation/Thesis
T1 Goal reasoning for autonomous agents using automated planning
A1 Pozanco Lancho, Alberto
AB Automated planning deals with the task of finding a sequence of actions, namelya plan, which achieves a goal from a given initial state. Most planning researchconsider goals are provided by a external user, and agents just have to find aplan to achieve them. However, there exist many real world domains whereagents should not only reason about their actions but also about their goals,generating new ones or changing them according to the perceived environment.In this thesis we aim at broadening the goal reasoning capabilities of planningbasedagents, both when acting in isolation and when operating in the sameenvironment as other agents.In single-agent settings, we firstly explore a special type of planning taskswhere we aim at discovering states that fulfill certain cost-based requirementswith respect to a given set of goals. By computing these states, agents are ableto solve interesting tasks such as find escape plans that move agents in to safeplaces, hide their true goal to a potential observer, or anticipate dynamically arrivinggoals. We also show how learning the environment’s dynamics may helpagents to solve some of these tasks. Experimental results show that these statescan be quickly found in practice, making agents able to solve new planningtasks and helping them in solving some existing ones.In multi-agent settings, we study the automated generation of goals based onother agents’ behavior. We focus on competitive scenarios, where we are interestedin computing counterplans that prevent opponents from achieving theirgoals. We frame these tasks as counterplanning, providing theoretical propertiesof the counterplans that solve them. We also show how agents can benefitfrom computing some of the states we propose in the single-agent setting toanticipate their opponent’s movements, thus increasing the odds of blockingthem. Experimental results show how counterplans can be found in differentenvironments ranging from competitive planning domains to real-time strategygames.
YR 2021
FD 2021-12
LK https://hdl.handle.net/10016/35328
UL https://hdl.handle.net/10016/35328
LA eng
NO Mención Internacional en el título de doctor
DS e-Archivo
RD 1 sept. 2024