Unreal Home is a simulation platform designed to carry out the generation of photorealistic synthetic datasets of
human daily basis interactions in virtual environments.
It was developed in collaboration with the 3D Perception Lab (3DPL) group from
the Department of Information Technology and Computing (DTIC) of the
University of Alicante.
GITHUB REPOSITORY
UnrealHome is a simulation platform designed to carry out the generation of visual and photorealistic
synthetic datasets from human interactions with objects in virtual home environments. It has been developed
using the Unreal Engine 4 game engine, and is inspired by other similar projects such as
UnrealRox,
VirtualHome
and ElderSim.
This software is composed of two types of actors (Avatars and Objects), which are directly involved in the execution
of the simulations, and a total of three components or subsystems responsible for carrying out all the necessary threads
that ensure the correct behavior of UnrealHome’s internal logic.
These threads include the playback and interpolation of animations managed by the animation subsystem, the movement of
the avatars through the scene carried out by the navigation and pathfinding subsystem, and the grabs of objects made by
the avatars, which are managed by the grip subsystem.
Finally, this project has also sought to ensure that the generation of visual synthetic datasets becomes more interactive
and intuitive. Therefore, UnrealHome makes it possible to use Point and Click mechanics in real time during the simulations
with the aim of sending instructions to the avatars of the scene so that they carry out new interactions defined by said
instructions.
• Extended UE4 source code to allow dynamic animation sampling in the BlendSpace1D assets.
• Developed a grip system to allow the avatars to grab objects with their hands.
• Implemented Point and Click mechanics for a more interactive use of the software.
• Implemented realistic physics simulation and collision management.
• Coded the avatar AI system (Pathfinding + Finite State Machines).
• Coded and designed an intuitive and simple UI.