Print Email Facebook Twitter Generative CoLearn: steering and cost prediction with generative adversarial nets in kinodynamic RRT Title Generative CoLearn: steering and cost prediction with generative adversarial nets in kinodynamic RRT Author Tsutsunava, Nick (TU Delft Mechanical, Maritime and Materials Engineering; TU Delft Biorobotics) Contributor Wolfslag, Wouter (mentor) Hernandez Corbato, Carlos (mentor) Wisse, Martijn (graduation committee) Kober, Jens (graduation committee) de Bruin, Tim (graduation committee) Degree granting institution Delft University of Technology Date 2018-10-05 Abstract Kinodynamic planning is motion planning in state space and aims to satisfy kinematic and dynamic constraints. To reduce its computational cost, a popular approach is to use sampling based methods such as RRT with off-line machine learning for estimating the steering cost and inputs. However, scalability and robustness are still open challenges in these type of Learning-RRT algorithms. We propose the use of generative adversarial networks (GAN) for learning of the steering cost and inputs. Furthermore, a novel data generation method is introduced, which is easy to learn and, in terms of parameter count, scales linearly to higher degrees of freedom. In our experiments, we show that the GAN has excellent generalisation capabilities, resulting in a considerable improvement in performance compared to the state-of-the-art. Consequently, we show that our method can scale to a planar arm and is robust to data dimensionality. Subject Deep LearningPath PlanningOptimal Control To reference this document use: http://resolver.tudelft.nl/uuid:7953081a-1cf1-4e4b-8ca4-87908ffcfac5 Embargo date 2019-01-31 Part of collection Student theses Document type master thesis Rights © 2018 Nick Tsutsunava Files PDF generative_colearn_appendix.pdf 2.16 MB PDF generative_colearn_thesis_cover.pdf 2.92 MB Close viewer /islandora/object/uuid:7953081a-1cf1-4e4b-8ca4-87908ffcfac5/datastream/OBJ1/view