Treffer: TrafficFlow
Weitere Informationen
This Unity package contains work in progress on developing a heterogeneous, multiagent system that simulates traffic flow in the context of Non-Player Characters for game design. A Reinforcement Learning approch was equipped. To you use our work please cite the following paper as a reference: S. Makri and P. Charalambous, "Towards a multi-agent non-player character road network: a Reinforcement Learning approach", 2021 IEEE Conference on Games (CoG), 2021, https://ieee-cog.org/2021/assets/papers/paper_280.pdf. Installation Prerequisites The TrafficFlow package is implemented in the Unity game engine and utilises the ML-Agents Toolkit. Thus, installation of the following is a prerequisite: Install Unity (2018.4 or later) Install Python (3.6.1 or higher) Clone the ml-agents repository (release_13). Install the mlagents Python package (0.24.0) You can refer to the ml-agents repository documentation for more details. We recomend that you follow the local installation for development instructions when installing the mlagents Python package. NOTE: You may wish to omit installing the com.unity.ml-agents and com.unity.ml-agents.extensions Unity packages for the time being. We will come back to this later. Download the Trafficflow repository Clone this repository. git clone --branch v1.0 https://github.com/CYENS/TrafficFlow.git You then need to edit the ./Packages/manifest.json and .Packages/packages-lock.json files as follows. In line 8 of ./Packages/manifest.json, replace with the path where you installed the ml-agents repository. In line 82 of ./Packages/manifest.json, replace with the path where you installed the ml-agents repository. Installing the ML-Agents Unity package Open the TrafficFlow project in Unity. You should check that the following Unity packages are listed in the Package Manager. com.unity.ml-agents Unity package (1.8.0-preview) com.unity.ml-agents.extensions Unity package (0.1.0-preview) You can also install these mannually by adding the local com.unity.ml-agents and com.unity.ml-agents.extensions ...