Continuous Value Iteration (CVI) Reinforcement Learning and Imaginary Experience Replay (IER) for learning multi-goal, continuous action and state space controllers
| Autor: | Michael Spranger, Andreas Gerken |
|---|---|
| Rok vydání: | 2019 |
| Předmět: |
Computer Science::Machine Learning
FOS: Computer and information sciences 0209 industrial biotechnology Robot kinematics Computer Science - Artificial Intelligence business.industry Generalization Computer science 02 engineering and technology 010501 environmental sciences 01 natural sciences Nondeterministic algorithm 020901 industrial engineering & automation Artificial Intelligence (cs.AI) Task analysis Benchmark (computing) Robot Reinforcement learning State space Artificial intelligence Markov decision process business 0105 earth and related environmental sciences |
| Zdroj: | ICRA |
| DOI: | 10.48550/arxiv.1908.10255 |
| Popis: | This paper presents a novel model-free Reinforcement Learning algorithm for learning behavior in continuous action, state, and goal spaces. The algorithm approximates optimal value functions using non-parametric estimators. It is able to efficiently learn to reach multiple arbitrary goals in deterministic and nondeterministic environments. To improve generalization in the goal space, we propose a novel sample augmentation technique. Using these methods, robots learn faster and overall better controllers. We benchmark the proposed algorithms using simulation and a real-world voltage controlled robot that learns to maneuver in a non-observable Cartesian task space. Comment: Published in 2019 International Conference on Robotics and Automation (ICRA) 20-24 May 2019 |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|