Abstract: Planned operation paths can often result in the relatively low field turning efficiency of the machinery, particularly in the current path planning for unmanned oilseed rape sowing with intelligent agricultural machinery. Furthermore, the path planning cannot consider the fuel consumption during turning and the mid-operation replenishment of seed and fertilizer boxes, leading to low overall operation economy. In this study, an optimal path was proposed to consider the mid-operation resupply (OPCMR). Firstly, the fish-tail and U-shaped turning models were often adopted to alter the rows in rectangular fields. Different specifications were further classified according to the number of rows between operations. The fuel consumption curves of the machinery were measured under different states during turning. Secondly, the two turning models were constructed using the maximum turning radius and working width of the machinery. A turning cost function was also established with the turning time and fuel consumption of the machinery. The turning models were taken as the constraints. The slow convergence and easy trapping were confined to the local optimal solutions of the conventional ant colony algorithm (ant colony optimization, ACO). The conventional ACO algorithm was integrated with the simulated annealing algorithm and the heuristic factor of the ACO algorithm. An improved ant colony algorithm was obtained (improved ant colony optimization, IACO). Thirdly, the optimal operation path was generated with the least turning time and fuel consumption using the turning cost function and the IACO algorithm. The function was added to manually arrange the field entrance in the path planning, thus considering the uncertainty of the field entrance position. The first operation of the planned path started from the position closest to the field entrance. Finally, an autonomous seed and fertilizer replenishment control system was proposed for intelligent agricultural machinery. The key parameters were determined, such as the capacity of the seed and the fertilizer box of the combined seeding and fertilizing machine, as well as the per mu sowing and fertilizing amount in the field operation. Furthermore, the specific path interruption points were generated on the optimal operation path. These points were connected with the pre-set seed. The fertilizer replenishment stations were generated with the specific seed and fertilizer replenishment paths. Once the seed and fertilizer of the machinery were insufficient, the seed and fertilizer replenishment station was shifted along the path for replenishment, and then returned along this path to continue the operation. The results show that the IACO algorithm reduced the number of iterations by up to 56.9%, compared with the ACO algorithm. The minimum turning cost value was reduced by up to 2.82%, effectively avoiding the slow convergence and easy trapping in local optimal solutions of the ACO algorithm. The seed and fertilizer replenishment path tracking experiments show that the average distance between the machinery parking point and the seed and fertilizer replenishment station was no more than 0.5 m, and the average initial deviation was no more than 0.03 m when the machinery entered the next operation row. The seed and fertilizer replenishment control effectively realized the seed and fertilizer replenishment task without interrupting the operation of the next row. Compared with the conventional comb-shaped operation path, better performance was achieved in the unmanned agricultural machinery along the OPCMR path, with a 17.5% increase in the turning efficiency, a 9.8% reduction in turning fuel consumption, and a 94.7% reduction in human resource input after operation. The path planning effectively improved the operation efficiency to reduce the resource input. This finding can provide the technical support to construct the unmanned oilseed rape farms.