Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!sdd.hp.com!usc!ucla-cs!oahu.cs.ucla.edu!alexis From: alexis@oahu.cs.ucla.edu (Alexis Wieland) Newsgroups: comp.ai.neural-nets Subject: Pole Balancing Equations Message-ID: <1991Feb3.185222.15543@cs.ucla.edu> Date: 3 Feb 91 18:52:22 GMT Sender: news@cs.ucla.edu (Shemp News Account) Organization: UCLA Computer Science Department Lines: 20 Originator: alexis@oahu.cs.ucla.edu Nntp-Posting-Host: oahu.cs.ucla.edu > Has anyone implemented the pole balancing simulator whoses equations are > given in Barto and Sutton "Neuronlike Adaptive Elements That Can Solve ... > Does anyone know what is wrong with the equations?? ... Without checking your implementation, Barto, Sutton, and Anderson have gravity going the wrong way, try changing the sign of 'g'. BTW, the equations for many poles on one cart and for jointed poles are in "Evolving Controls for Unstable Systems", A. Wieland, Connectionist Models: Proceedings of the 1990 Summer School, Morgan Kaufmann, 1991, pp 91-102. Warning: J. Schmidhuber (for yet a different approach to pole balancing see his paper "Making the World Differentiable: On Using Supervised Learning Fully Recurrent Neural Networks for Dynamic Reinforcement Learning and Planning in Non-Stationary Environments", TR: FKI-126-90, Technische Universitat Munchen, West Germany) has told me that I made the same error - alexis wieland