|
(See ISLA for the new IAS website. Most maintenance activities on this old website have stopped.)
Reinforcement learning
|
Reinforcement learning and neurocontrol
|
Objective
In a joint project with the Control Group from the Delft University of Technology we investigate Reinforcement learning (RL) for control applications.
The objective is to investigate RL in the continuous state domain and to
study the applicability for real control problems.
Research group members
dr. ir. S. ten Hagen
dr. ir. B. Kröse
Funding
Stichting Technische Wetenschappen (STW)
Research Achievements
We investigated how much exploration (exitation) we have to add to the
control signal in order to get good learning performance. For the LQR
framework we found that conventional system identification approach needed
less exploration.
For non-linear systems we developed `Neural-Q' learning.
More information
Link to personnal page
|
