Integrating Neuro-Symbolic AI for Object Recognition and Robotic Task Planning

Humans and robots both learn from their environment and build knowledge through experience. However, unlike robots, humans rely on symbolic reasoning to plan future actions, allowing them to go beyond purely reactive behavior. NeuroSymbolic AI seeks to bridge this gap by enabling robots to reason and plan using symbolic structures, in a manner similar to how humans approach goal-directed tasks. This thesis investigates the application of NeuroSymbolic AI in robotic manipulation. In particular, it presents an implementation using the TIAGo robot and the DeepSym framework for path planning and object manipulation. The objective is for the robot to reach the position of a target object by learning from its environment in an unsupervised manner, demonstrating how symbolic reasoning can be effectively integrated into real-world robotic tasks.