Rectangle expansion A* pathfinding for grid maps

Search speed, quality of resulting paths and the cost of pre-processing are the principle evaluation metrics of a pathfinding algorithm. In this paper, a new algorithm for grid-based maps, rectangle expansion A＊ （REA＊）, is presented that improves the performance of A＊ significantly. REA＊ explores maps in units of unblocked rectangles. All unnecessary points inside the rectangles are pruned and boundaries of the rectangles （instead of individual points within those boundaries） are used as search nodes. This makes the algorithm plot fewer points and have a much shorter open list than A＊. REA＊ returns jump and grid-optimal path points, but since the line of sight between jump points is protected by the unblocked rectangles, the resulting path of REA＂ is usually better than grid-optimal. The algorithm is entirely online and requires no offline pre-processing. Experimental results for typical benchmark problem sets show that REA＊ can speed up a highly optimized A＊ by an order of magnitude and more while preserving completeness and optimality. This new algorithm is competitive with other highly successful variants of A＊.

An improved master-apprentice evolutionary algorithm for minimum independent dominating set problem

The minimum independent dominance set(MIDS)problem is an important version of the dominating set with some other applications.In this work,we present an improved master-apprentice evolutionary algorithm for solving the MIDS problem based on a path-breaking strategy called MAE-PB.The proposed MAE-PB algorithm combines a construction function for the initial solution generation and candidate solution restarting.It is a multiple neighborhood-based local search algorithm that improves the quality of the solution using a path-breaking strategy for solution recombination based on master and apprentice solutions and a perturbation strategy for disturbing the solution when the algorithm cannot improve the solution quality within a certain number of steps.We show the competitiveness of the MAE-PB algorithm by presenting the computational results on classical benchmarks from the literature and a suite of massive graphs from real-world applications.The results show that the MAE-PB algorithm achieves high performance.In particular,for the classical benchmarks,the MAE-PB algorithm obtains the best-known results for seven instances,whereas for several massive graphs,it improves the best-known results for 62 instances.We investigate the proposed key ingredients to determine their impact on the performance of the proposed algorithm.