In computer graphics, physically-based global illumination algorithms such as photon-mapping have a linear progression between complexity and quality. To a given quality, rendering time scale linearly with computer performances. With Moore's law call in question and increasing demand in quality thoose algorithms need more and more optimisations. Existing optimisations such as irradiance caching are themselves very linear in performance gain. We propose a new optimisation method for the photon-mapping global illumination algorithm. We used structures borrowed from Artificial Intelligence such as autonomous agents to reduce computing time of the photon casting phase and rendering phase of the algorithm. Our structure called Autonomous Lighting Agents starts by using an agent-based scene discovery algorithm and is later used to make decisions during rendering inducing less photons beeing casted and smaller rendering times.
