Using field-programmable gate arrays for learning non player characters
Christopher Cichiwskyj, Gregor Schiele · 2016
In this paper we present ongoing work on how to use Field-Programmable Gate Arrays to increase the number of concurrent non player characters in large scale interactive virtual worlds. We employ reinforcement learning combined with artificial neural networks to allow the simulated characters to learn from previous engagements with players. Our simulations show achievable performance gains of several orders of magnitude compared to a CPU-based solution.