Ögrenme güdümlü olasılıksal planlama

Robots should avoid potential failure situations to improve their performances. The failures which the robot has already experienced in its previous action executions can be used in an adaptive planning strategy to reduce potential failures in the future. Robots need to build and use their experience for achieving this objective. In this thesis, learning and learning-guided planning methods are proposed to address this problem. An experimental learning process using Inductive Logic Programming (ILP) and a probabilistic planning method that uses the experience gained by learning are integrated for improving task execution performance. The solutions are analyzed on a case study with an autonomous mobile robot in a multi-object manipulation domain where the objective is maximizing the number of collected objects while avoiding potential failures using experience. Obtained results indicate that the robot using the adaptive planning strategy ensures safety in task execution and maximizes the probability of success.