Multi-task Scheduling of Multiple Agricultural Machinery via Reinforcement Learning and Genetic Algorithm

Scheduling of agricultural machinery is a crucial aspect of smart agriculture. This paper addresses the multi-task scheduling problem for multiple agricultural machinery, which is challenging and belongs to NP-hard. However, traditional intelligent algorithms solve this problem by fixing their key parameters in a preset way. This may cause efficiency and quality of the solution to be unable to meet the production requirements. Therefore, this paper proposes a self-learning algorithm (SLA), which combines the Sarsa algorithm and the genetic algorithm. Firstly, a combined model is constructed. Secondly, the Markov decision process's state, action, reward, and strategy are designed. Finally, the performance of SLA in solving multi-tasks of multiple agricultural machinery is compared to baselines using eighteen groups of instances with different scales. The experimental results show that the proposed SLA significantly outperforms baselines in solving multi-task scheduling of multiple agricultural machinery. This finding can provide a reasonable scheduling scheme for the production of large-scale unmanned farms.