Abstract: To improve the path search efficiency and path optimization of mine rescue UAVs in environments with narrow passages and dense, complex obstacles, a path planning method based on the improved Artificial Jellyfish Search (IJS) algorithm was proposed. The Artificial Jellyfish Search (JS) algorithm was combined with the Logistic chaotic mapping to update pheromones and avoid falling into local optima. A Gaussian mutation function was applied to reduce the number of poor-quality individuals in the population. A Lévy flight disturbance strategy was introduced to optimize the position update formulas during the phases of drifting with ocean currents (global search) and tracking food sources (local search), thereby improving the efficiency of UAV path planning. UAV path planning simulation experiments showed that, when the obstacle ratio was 14.56%, compared with the Genetic Algorithm (GA), Particle Swarm Optimization (PSO) algorithm, and JS algorithm, the IJS algorithm reduced the path planning time by 72.27%, 66.12%, and 70.87%, respectively; shortened the path length by 2.67%, 3.95%, and 1.36%, respectively; and reduced the number of turning points by 47.37%, 50%, and 28.57%, respectively. When the obstacle ratio was 32.20%, compared with GA and PSO, the IJS algorithm reduced the planning time by 62.50% and 55.61%, shortened the path length by 4.03% and 4.03%, and reduced the number of turning points by 15.38% and 18.52%, respectively. Compared with the JS algorithm, although the path length increased by 3.89%, the planning time was reduced by 57.32%, and the number of turning points decreased by 8.33%. A post-disaster underground tunnel experimental platform was built to conduct UAV path planning experiments. The results showed that, compared with the GA, PSO, and JS algorithms, the IJS algorithm reduced the path planning time by 60.77%, 58.70%, and 51.52%, respectively; shortened the path length by 9.62%, 7.58%, and 7.50%, respectively; and reduced the number of turning points by 40%, 30.77%, and 25%, respectively. These results verified that the IJS algorithm possesses better path optimization capability and computational efficiency than the compared algorithms in complex environments.