1: \begin{abstract}

2:   This paper addresses the challenges of real-time, large-scale, and near-optimal multi-agent pathfinding (MAPF) through enhancements to the recently proposed \lacamstar algorithm.

3:   \lacamstar is a scalable search-based algorithm that guarantees the eventual finding of optimal solutions for cumulative transition costs.

4:   While it has demonstrated remarkable planning success rates, surpassing various state-of-the-art MAPF methods, its initial solution quality is far from optimal, and its convergence speed to the optimum is slow.

5:   To overcome these limitations, this paper introduces several improvement techniques, partly drawing inspiration from other MAPF methods.

6:   We provide empirical evidence that the fusion of these techniques significantly improves the solution quality of \lacamstar, thus further pushing the boundaries of MAPF algorithms.

7: \end{abstract}

8: