Multi-objective optimization of the operation node sequence of rice harvesters

Operation node sequence can dominate the operation distance and time of harvesters. Especially, the capacity constraints are limited to the size of the grain tank. In this study, a multi-objective optimization model was presented with the variables of operation node sequence, in order to reduce the operation distance and time during rice harvesting. The operation distance consisted of rice harvesting and non-working distance. The operation time included the time for rice harvesting and the time spent on non-working activities, such as returning to the unloading point. The constraints were included in the multi-objective optimization model, such as the capacity of the grain tank and repeating path. The effectiveness of the improved model was then validated after optimization. Each row of rice was required to be fully harvested at once during rice harvesting. Once the harvester reached the maximum capacity of the grain tank or the remaining capacity of the grain tank could not accommodate the next row of rice, the harvester was returned to the unloading point. Only one unloading point was found in each field. Among them, the location was the same as the position where the harvester entered the field. An improved ant colony optimization (IACO) algorithm was developed to solve this multi-objective optimization model using the traditional ant colony optimization (ACO). Each ant represented a potential solution in the IACO algorithm. An optimal or near-optimal solution was converged after iterative exploration and pheromone updating. The selection principle of the IACO was combined with the deterministic and stochastic analysis, in order to avoid stagnation during searching. The solution was selected with the highest probability using the deterministic principle. While the roulette wheel was employed to introduce the randomness in the stochastic principle. This hybrid approach also enabled the algorithm to escape the local optima for accelerated convergence over the short runtime. Additionally, the pheromone update strategy was refined to only update the pheromones on the best-performing path of ant. According to the TOPSIS (technique for order preference by similarity to an ideal solution) evaluation, the optimal weighting scheme was achieved, where the weight of operation distance was 0.6 and the weight of operation time was 0.4. The parameters of the IACO algorithm were determined after sensitivity analysis of each factor. The number of ants was 120, the pheromone factor was 1, the heuristic function factor was 3, and the pheromone volatilization factor was 0.3. A simulation was conducted to validate the improved model using the traditional ACO and the IACO. The results show that the IACO optimization reduced the operation distance and its standard deviation by 5.47% and 71%, respectively. Moreover, the IACO also decreased the operation time and its standard deviation by 5.29% and 47%, respectively. The running time and its standard deviation decreased by 12.25% and 37.07%, respectively. The superiority and stability of IACO were highlighted after optimization. Field experiments further demonstrate that the multi-objective optimization reduced the operation distance and time by 9.5% and 9.1%, respectively, in the trapezoidal field, compared with the traditional ox-plowing. In rectangular fields, the operation distance and time were reduced by 2.6% and 6.9%, respectively. The multi-objective optimization and the IACO algorithm were effective for the path planning of harvesters. The finding can also provide a better solution to intelligent multi-machine collaborative harvesting.