1. What is the purpose of the new web-based system for MAPF?
The new web-based system for MAPF aims to track and validate results on standardized benchmarks, map the current pareto-frontier, identify strengths and challenges in the area, and track progress over time. It also helps lower the barrier of entry for new research on MAPF.
read more
2. What are the three types of algorithms studied by the research community in MAPF benchmarks?
The three types of algorithms studied by the research community in MAPF benchmarks are: (i) Optimal Algorithms, which focus on finding exact optimal solutions by starting from a lower-bound and progressively pushing it until a provably optimal feasible solution is found. (ii) Bounded Suboptimal Algorithms, which find suboptimal solutions within theoretical guarantees by exploring lower-bounds and feasible solutions simultaneously, returning a solution that meets certain suboptimality. (iii) Unbounded Suboptimal Algorithms, which focus on finding feasible solutions quickly and improving them given sufficient time. These algorithms do not have theoretical guarantees for suboptimality. Our system tracks the progress of these algorithms by recording the best lower-bound value and the best solution cost, along with additional metadata such as the algorithm name, authors, publication references, and links to implementations.
read more
3. How does the system facilitate researcher participation?
The system allows researchers to participate by submitting their algorithms/results, establishing the state-of-the-art together. It makes all collected results publicly available for download and provides tools for automatic algorithm comparison across every level. The principal evaluation criteria include the number of instances closed, solved, best lower-bound achieved, and best solution reported. These criteria help summarize the state-of-the-art for each algorithm type.
read more