Title :
Randomly Sampling Actions In Dynamic Programming
Author :
Atkeson, Christopher G.
Author_Institution :
Robotics Inst., Carnegie Mellon Univ., Pittsburgh, PA
Abstract :
We describe an approach towards reducing the curse of dimensionality for deterministic dynamic programming with continuous actions by randomly sampling actions while computing a steady state value function and policy. This approach results in globally optimized actions, without searching over a discretized multidimensional grid. We present results on finding time invariant control laws for two, four, and six dimensional deterministic swing up problems with up to 480 million discretized states
Keywords :
dynamic programming; random processes; sampling methods; deterministic dynamic programming; randomly sampling; time invariant control laws; value function; Computational efficiency; Cost function; Dynamic programming; Interpolation; Learning; Multidimensional systems; Robots; Sampling methods; Steady-state; USA Councils;
Conference_Titel :
Approximate Dynamic Programming and Reinforcement Learning, 2007. ADPRL 2007. IEEE International Symposium on
Conference_Location :
Honolulu, HI
Print_ISBN :
1-4244-0706-0
DOI :
10.1109/ADPRL.2007.368187
