Tubitak #111E286 Project on automated planning and learning methods for autonomous mobile robots

Summary:

In this project, we propose an integrated planning, execution and learning framework for cognitive robots to ensure their safe action execution. This framework ensures that robots detect their failures in runtime and learn from real-world experimentation. Robots need to maintain a consistent world model for detecting failures in the environment. For this purpose, we designed a temporal scene interpretation unit that integrates different sensory modalities. Objects in the scene are recognized by using both color and depth information, and the unknown objects are segmented by using Euclidean clustering on the depth values. Action execution failures are detected by continually monitoring Metric Temporal Logic formulas. Robots gain experience on actions, objects in interest and their relations by using Inductive Logic Programming as the lifelong experimental learning method. Experience gained through learning is represented by first-order logic that is useful for reasoning and planning processes. This experience is used to guide future plans of robots. The performance of the framework is analysed on mobile robots and a robotic arm on object manipulation scenarios. The results reveal that the framework ensures that the scenes are interpreted correctly, the failures are detected with over 95% success rate on average, hypotheses framed for failure cases outperform other supervised learning methods with over 93% success rate on average, and the related learning-guided planning methods ensure safety in future tasks of the robot.

This research is funded by a grant from the Scientific and Technological Research Council of Turkey (TUBITAK), Grant No. 111E-286.

Project Team:

PI: Dr. Sanem Sariel

Researchers: Dr. Hulya Yalcin, Mustafa Ersen, Melis Kapotoglu, Melodi Deniz Ozturk, Sertac Karapinar, Cagatay Koc, Dogan Altan, Petek Yildiz, Burak Topal and Mehmet Biberci

Advisor: Dr. Muhittin Gokmen

Special thanks to: Yuksel Seker, Yilmaz Seker, Peyman Beyranvand, Naseem Al-Housani, Ercin Temel, Burak Sarigul, Mehmet Sarioglu, Ahmet Fehmi Ozcan, Gozde Ozcan, Tugba Yagiz, Abdullah Cihan Ak, Talha Colakoglu, Kutay Hekimoglu, Selen Sariel and our department’s faculty and staff.