Deep Reinforcement Learning for High Level Character Control

  • 2020-05-20 23:32:19
  • Caio Souza, Luiz Velho
  • 11


In this paper, we propose the use of traditional animations, heuristicbehavior and reinforcement learning in the creation of intelligent charactersfor computational media. The traditional animation and heuristic gives artisticcontrol over the behavior while the reinforcement learning adds generalization.The use case presented is a dog character with a high-level controller in a 3Denvironment which is built around the desired behaviors to be learned, such asfetching an item. As the development of the environment is the key forlearning, further analysis is conducted of how to build those learningenvironments, the effects of environment and agent modeling choices, trainingprocedures and generalization of the learned behavior. This analysis buildsinsight of the aforementioned factors and may serve as guide in the developmentof environments in general.


Quick Read (beta)

loading the full paper ...