Aiming to address the challenges of slow convergence, low search efficiency, and limited global search capability in traditional ant colony optimization (ACO) algorithms for agricultural robot path planning, a multi-strategy fusion improved ant colony algorithm (MSFIACO) was proposed. The MSFIACO integrated a multi-layer vision expansion search strategy and introduced a non-uniform pheromone initialization with an angular judgment mechanism. This mechanism assigned higher initial pheromone concentrations to paths closer to the straight line connecting the start and end points. A reward-punishment constraint was applied along the ants' movement direction to enhance search efficiency and precision. Additionally, an adaptive pseudo-random state transition strategy was introduced to balance global exploration and convergence speed, while a smoothing factor was used to reduce turning frequency. An adaptive heuristic information strategy was also proposed to ensure efficient convergence in the early stages and effective global search in the later stages. The pheromone update rule was refined to distinguish between superior and inferior solutions, focusing the search on the neighborhoods of promising paths. Experiments showed that MSFIACO significantly reduced path length, decreased turning points, accelerated convergence, reduced the number of path nodes, and exhibited strong robustness and adaptability, enhancing the path planning capability of agricultural robots in complex environments.