@PHDTHESIS{ 2020:671044036,
 	title = {Symbolic-geometric planning},
 	year = {2020},
 	url = "https://tede2.pucrs.br/tede2/handle/tede/10964",
 	abstract = "Classical planners, with actions described with preconditions and effects, create a way to operate on purely symbolic models in order to find plans to reach an agent’s goals. Plans found by classical planners often lack the geometric details required to solve motion problems, such as grasping an object or avoiding narrow passages. Such details include robot dimensions and object properties. Conversely, motion planners consider only physical details, not symbolic goals or parts of the state. Both symbolic and motion planners are required to solve certain problems, but relations between symbolic and geometric data must be shared to avoid replanning large portions of the search-space. In order to tackle this problem, we need a symbolic-geometric planner to share data and limit possible values as planning progresses, to minimize memory usage and planning time. Different planning algorithms for this type of hybrid planning have been developed to address the problem of combining geometric and symbolic planning by using off-the-shelf planners or new implementations. Most hybrid planners share information between the parts using a fixed set of symbols, which limits the amount of information that can be shared. By generating such symbols during planning for the symbolic part and relating them to external complex objects (containers, structs, instances) in the geometric part, it is possible to simplify the symbolic description while exploiting complex structures and functions already available in external libraries, such as the ones used by simulators. The contributions of this work include the definition of semantic attachments, as a mechanism to share information between symbolic and external/geometrical parts for Hierarchical Task Network (HTN) planning, a symbol to object table to keep external details hidden from the symbolic part, while able to compute with external objects using functions and semantic attachments, and a precondition reordering algorithm to improve planning time.",
 	publisher = {Pontifícia Universidade Católica do Rio Grande do Sul},
 	scholl = {Programa de Pós-Graduação em Ciência da Computação},
 	note = {Escola Politécnica}
}