Enhancing Greedy Best-First Search with Dynamic Heuristic for Puzzle Solving
Tina Babu, Deepika Nayak, Shiva Kumar, Shweta, O Nishitha, Saket Mishra · 2025
Because the GBFS algorithm uses a heuristic-driven methodology giving priority to the most promising nodes, this algorithm is frequently used in the solving of puzzles and pathfinding problems. However, when the heuristic function can-not provide apt guidance, the GBFS algorithm often encounters problems such as choosing less-than-ideal routes or becomes stuck in local minima. For especially complex problems requiring complex state evaluation, this may imply longer search times and generally lower-quality answers. A dynamic GBFS system with adaptive heuristic modification is presented to overcome these drawbacks. Contrasted with traditional GBFS that relies on a fixed heuristic at each step of the search process, this proposed algorithm modifies the heuristic dynamically on-line based on real-time feedback from its current search operations. By judging the effectiveness of previously known routes, this adaptive heuristic improves Classic puzzles such as the 8-puzzle, 15-puzzle, and sliding tile puzzles are considered to test the new GBFS. Experimental results state that the proposed method significantly outperformed the basic GBFS in search time and quality of solutions. Such robustness explains how this algorithm could potentially work well in scenarios where heuristic correctness is essential by remaining insensitive to varying complexity. This research is therefore groundbreaking, for it proves the benefits of dynamic heuristic adjustment in making puzzle-solving techniques even more effective and dependable.