An Online Interactive Demo
PRIMAL GRID is an innovative online interactive web application developed by Insaf, as part of his Final Year Project with the guidance from his Project Supervisor, Professor Sartoretti. The application builds upon the success of PRIMAL, a research paper by Professor Sartoretti and his team of researchers. PRIMAL proposes a machine learning model for Multi-Agent Pathfinding (MAPF) applications that relies on distributed reinforcement learning. The paper has gained widespread attention from researchers and practitioners interested in exploring the capabilities of machine learning models for MAPF, being the first work to break the 1000-robot barrier while yielding high-quality paths.
PRIMAL GRID provides a powerful tool for researchers and developers to explore the capabilities of machine learning models in an accessible and user-friendly manner. Since its launch on March 6, 2023, PRIMAL GRID has attracted users from around the world and yielded promising results. The application enables researchers to create custom MAPF instances on square grid worlds by placing obstacles, agents, and corresponding goals. Users can then run a trained PRIMAL model in real-time and view agents as they plan individual collision-free paths to their goals. The application dynamically renders a real-time visualization of each agent's path to the goal, improving the visualization of machine learning models and the overall workflow of MAPF researchers. PRIMAL GRID features an intuitive user interface and leverages cutting-edge technologies such as ReactJS, TensorFlowJS, and NodeJS. The system architecture follows a client-server model, with the client-side implemented in ReactJS and the server-side implemented in ExpressJS and NodeJS.