A Comparative Analysis of Informed and Uninformed Search Algorithms for the 8-Puzzle Problem with a Novel Hybrid Approach

Abstract

The 8-puzzle problem serves as a fundamental benchmark for evaluating search algorithm performance in artificial intelligence applications. This paper presents a comprehensive comparative analysis of informed and uninformed search algorithms for solving the 8-puzzle problem, including Breadth-First Search (BFS), Depth-First Search (DFS), Greedy Best-First Search, and A* Search. Additionally, we introduce a novel hybrid approach called Hybrid-ADFS (A-Driven Depth-First Search) that integrates the heuristic guidance of A* with the memory-efficient exploration of depth-limited DFS. The study evaluates algorithm performance across four critical dimensions: completeness, optimality, time complexity, and space complexity. Experimental results on ten carefully selected 8-puzzle instances demonstrate that while A* provides the most balanced performance with guaranteed optimal solutions, the proposed Hybrid-ADFS algorithm achieves competitive execution times with significantly reduced node expansion (9-75 nodes compared to 71-887 for A*) while maintaining complete problem-solving capability. BFS ensures optimality but suffers from prohibitive memory consumption (up to 49MB), while DFS offers excellent memory efficiency but lacks reliability, failing on 20% of test cases. Greedy Best-First Search achieves superior speed but produces severely suboptimal solutions with path lengths up to 346% longer than optimal. The Hybrid-ADFS algorithm demonstrates promising characteristics for resource-constrained environments, successfully solving all test instances with execution times ranging from 0.013 to 0.095 seconds and memory usage between 185-1,112 KB. These findings provide valuable insights for algorithm selection in practical AI applications and establish the potential of hybrid approaches for balancing optimality, efficiency, and resource management in search problems.