English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Recent advances in Deep Reinforcement Learning (DRL) have largely focused onimproving the performance of agents with the aim of replacing humans in knownand well-defined environments. The use of these techniques as a game designtool for video game production, where the aim is instead to create Non-PlayerCharacter (NPC) behaviors, has received relatively little attention untilrecently. Turn-based strategy games like Roguelikes, for example, presentunique challenges to DRL. In particular, the categorical nature of theircomplex game state, composed of many entities with different attributes,requires agents able to learn how to compare and prioritize these entities.Moreover, this complexity often leads to agents that overfit to states seenduring training and that are unable to generalize in the face of design changesmade during development. In this paper we propose two network architectureswhich, when combined with a \emph{procedural loot generation} system, are ableto better handle complex categorical state spaces and to mitigate the need forretraining forced by design decisions. The first is based on a dense embeddingof the categorical input space that abstracts the discrete observation modeland renders trained agents more able to generalize. The second proposedarchitecture is more general and is based on a Transformer network able toreason relationally about input and input attributes. Our experimentalevaluation demonstrates that new agents have better adaptation capacity withrespect to a baseline architecture, making this framework more robust todynamic gameplay changes during development. Based on the results shown in thispaper, we believe that these solutions represent a step forward towards makingDRL more accessible to the gaming industry.

Deep Policy Networks for NPC Behaviors that Adapt to Changing Design Parameters in Roguelike Games