1. What is the Multi-Agent Path Finding (MAPF) problem?
The Multi-Agent Path Finding (MAPF) problem is a well-known combinatorial problem that has been proven to be NP-hard, making it difficult to compute an optimal solution. It involves finding paths for multiple agents in a given environment, considering their interactions and avoiding collisions. Optimal algorithms exist but tend to be slow and poorly scale with large environments. Subsolvers are often used to provide sub-optimal solutions, either bounded or unbounded, to meet scalability and efficiency requirements. However, these algorithms focus on the algorithmic aspect and overlook other factors like network topology, goal sets, and path feasibility. The goal is to develop a system that considers these aspects for better short-term and long-term support in solving the MAPF problem.
read more
2. What are the two possibilities exploited by the system?
The system exploits two possibilities: 1) Noticing that the two goals are near the wrong agents, simplifying the problem and leading to better solutions. 2) Creating one-way corridors in the map topology to prevent agent conflicts, based on a database of past goals and evolutionary algorithms. This helps optimize routes and improve results. The system aims to create a holistic framework for multi-goal reasoning in real-life scenarios, considering different aspects of the problem and optimizing agent movements.
read more