Have a look at a research project that discusses the opportunities for deep learning in character animation and control.
The project is a part of Sebastian Starke's Ph.D. research at the University of Edinburgh in the School of Informatics, supervised by Taku Komura. The researcher developed a modular and stable framework for data-driven character animation, including data processing, network training and runtime control. Starke wanted to examine the advantages of using neural networks for animating biped locomotion, quadruped locomotion, and character-scene interactions with objects and the environment.
"Animating characters can be an easy or difficult task - interacting with objects is one of the latter. In this research, we present the Neural State Machine, a data-driven deep learning framework for character-scene interactions. The difficulty in such animations is that they require complex planning of periodic as well as aperiodic movements to complete a given task," states abstract.
The system is capable of synthesizing different movements and scene interactions from motion capture data, which allows end-users to seamlessly control the character in real-time from simple control commands. It gives the ability to set up a number of movements, "including locomotion, sitting on chairs, carrying boxes, opening doors and avoiding obstacles, all from a single model," noted Starke.
You can learn more about the project here.