Ongoing: Reinforcement Learning Robots Using LSTM
Robots that learn complex behavior, based on rewards
Behaviors that are hard to program, e.g. navigation in offices, object recognition and manipulation
Collect data from robot, learn controller in simulation, and fine tune again on real robot.
Exploit CSEM visual sensors
Bram Bakker, IDSIA Postdoc
Back to J. Schmidhuber's Recurrent neural network page