Robotic manipulation with pick and place task constructors

Abstract

Robotic systems operating in real-world environments are required to perform complex tasks that combine high-level decision making with low-level motion execution. Addressing these requirements calls for approaches that integrate symbolic task planning, geometric motion planning, and execution control. This thesis presents a framework for generating and executing Behavior Trees from a Task and Motion Planning (TAMP) problem, enabling structured and reactive execution of manipulation tasks. The proposed framework integrates symbolic task planning, motion planning, and execution using Behavior Trees with replanning capabilities. Task planning problems def inedinPlanningDomainDefinitionLanguage(PDDL)aresolvedtoobtainasequence of actions, from which the corresponding Behavior Tree is generated. Geometric informationabouttheenvironmentisthenextracted, andmotionplanningsolutionsforeach action are computed using the The Kautham Project tools. The nodes of the Behavior Tree responsible for controlling the robot’s movements access the computed trajectories to execute the actions within a Gazebo simulated environment. An execution module runs the generated Behavior Tree in a simulated Gazebo environment, while astate monitoring componentdetects inconsistencies between expected and perceived states and triggers replanning when necessary. The framework is validated through simulation of a pick-and-place manipulation task using the TIAGo Dual robot in a kitchen environment. In addition, a more specific approach for pick and place actions based on MoveIt 2 and the MoveIt Task Constructor is implemented. A comparative discussion highlights the trade-offs between the two approaches in terms of functionality, integration complexity, and access to planned trajectories. The results demonstrate that the proposed approach enables the systematic generation and execution of Behavior Trees from TAMP solutions, while supporting replanning in dynamic scenarios. The comparison of motion planning tools further emphasizes the importance of selecting planning frameworks according to application-specific requirements.