DCCI 2017 Abstracts

Full Papers
Paper Nr: 1

Lane Change Prediction in an Urban Area with an Echo State Network


Karoline Griesbach and Karl Heinz Hoffmann

Abstract: The prediction of a lane change and its integration in advanced driving assistance systems can reduce traffic accidents. The majority of earlier studies focused on lane change prediction with neural networks, Bayes networks, fuzzy systems or Hidden Markov models. In this paper an Echo State Network to distinguish lane change and no lane change sequences of urban areas is discussed. The preliminary results show that the distinction is successful.

Paper Nr: 2

Experience-based Competence Motivated Continuous Learning Framework


Paresh Dhakan, Kathryn Merrick, Inaki Rano and Nazmul Siddique

Abstract: Although multiple learning techniques exist to empower robots with different skills, open-ended lifelong learning is still an outstanding research problem in robotics. Open-ended learning would provide learning autonomy to robots such that they would not require any human intervention to learn. This Ph.D. project will design a continuous learning framework enabling robots to learn in an open-ended manner. Specifically, this project will focus on autonomous goal generation; combine it with new computational models of intrinsic motivation for goal selection and integrate it with reinforcement learning forming a continuous learning framework. This new framework will be validated on mobile robots enabling the robot to autonomously represent and learn complex skills in a self-motivated and open-ended way.