Adaptation technique for integrating genetic programming and reinforcement learning for real robots

Created by W.Langdon from gp-bibliography.bib Revision:1.4333

  author =       "Shotaro Kamio and Hitoshi Iba",
  title =        "Adaptation technique for integrating genetic
                 programming and reinforcement learning for real
  journal =      "IEEE Transactions on Evolutionary Computation",
  year =         "2005",
  volume =       "9",
  number =       "3",
  pages =        "318--333",
  month =        jun,
  ISSN =         "1089-778X",
  keywords =     "genetic algorithms, genetic programming, adaptive
                 systems, humanoid robots, learning (artificial
                 intelligence), legged locomotion, AIBO four-legged
                 robot, HOAP-1 humanoid robot, Q-learning method,
                 adaptation technique, box-moving task, reinforcement
                 learning, Adaptation evolutionary computation, box
                 moving, real robot, reinforcement learning (RL)",
  DOI =          "doi:10.1109/TEVC.2005.850290",
  size =         "16 pages",
  abstract =     "We propose an integrated technique of genetic
                 programming (GP) and reinforcement learning (RL) to
                 enable a real robot to adapt its actions to a real
                 environment. Our technique does not require a precise
                 simulator because learning is achieved through the real
                 robot. In addition, our technique makes it possible for
                 real robots to learn effective actions. Based on this
                 proposed technique, we acquire common programs, using
                 GP, which are applicable to various types of robots.
                 Through this acquired program, we execute RL in a real
                 robot. With our method, the robot can adapt to its own
                 operational characteristics and learn effective
                 actions. In this paper, we show experimental results
                 from two different robots: a four-legged robot AIBO and
                 a humanoid robot HOAP-1. We present results showing
                 that both effectively solved the box-moving task; the
                 end result demonstrates that our proposed technique
                 performs better than the traditional Q-learning
  notes =        "INSPEC Accession Number: 8465512

                 Graduate Sch. of Frontier Sci., Univ. of Tokyo, Chiba,

                 Reinforcement learning (RL) is outside the GP loop.
                 Q-table 168 or 238 states, RL ten hours or 6 hours. 6
                 way IF. GP run ten minutes. Target and box are colour
                 coded. Precise simulation of both robots is not
                 possible. p 330 GP {"}learning some general
                 knowledge...not limited to a particular robot{"}.",

Genetic Programming entries for Shotaro Kamio Hitoshi Iba