Carlos Florensa*, David Held*, Xinyang Geng*, Pieter Abbeel
In ICML, 2018.
Also in Deep Reinforcement Learning Symposium, NIPS 2017
We propose a method that allows an agent to automatically discover the range of tasks that it is capable of performing in its environment. We use a generator network to propose tasks for the agent to try to achieve, specified as sets of goal states.
Richard Zhang*, Jun-Yan Zhu*, Phillip Isola, Xinyang Geng, Angela S. Lin, Tianhe Yu, Alexei A. Efros
In SIGGRAPH 2017.
We propose a deep learning approach for user-guided image colorization. We system directly maps a grayscale image, along with sparse, local user “hints” to an output colorization with a deep convolutional neural network.
Marvin Zhang*, Xinyang Geng*, Jonathan Bruce*, Ken Caluwaerts, Massimo Vespignani, Vytas SunSpiral, Pieter Abbeel, Sergey Levine
In ICRA, 2017.
Also in Deep Reinforcement Learning Workshop, NIPS 2016.
We collaborated with NASA Ames to explore the challenges associated with learning locomotion strategies for tensegrity robots, a class of compliant robots that hold promise for future planetary exploration missions. We devised a novel extension of mirror descent guided policy search to learn locomotion gaits for the SUPERball tensegrity robot, both in simulationand on the physical robot.