Title:Learning by Viewing: Generating Test Inputs for Games by Integrating Human Gameplay Traces in Neuroevolution

Abstract:Although automated test generation is common in many programming domains, games still challenge test generators due to their heavy randomisation and hard-to-reach program states. Neuroevolution combined with search-based software testing principles has been shown to be a promising approach for testing games, but the co-evolutionary search for optimal network topologies and weights involves unreasonably long search durations. In this paper, we aim to improve the evolutionary search for game input generators by integrating knowledge about human gameplay behaviour. To this end, we propose a novel way of systematically recording human gameplay traces, and integrating these traces into the evolutionary search for networks using traditional gradient descent as a mutation operator. Experiments conducted on eight diverse Scratch games demonstrate that the proposed approach reduces the required search time from five hours down to only 52 minutes.