Title: Research on path planning for seedling transplanting using an improved Q-learning algorithm assisted by pheromones

Authors: Jianpu Cui; Ling Ren; Conghua Zhang; Xiangjian Zhu; Junping Liu; Wanyv Li; Xinchao Hu

Addresses: College of Mechanical and Electrical Engineering, Shihezi University, Heze – 274000, Shandong, China ' Shihezi University, Shihezi – 832000, Xinjiang, China ' Shihezi University, Dazhou – 635000, Sichuan, China ' Shihezi University, Wuhu – 241000, Anhui, China ' Shihezi University, Shihezi – 832000, Xinjiang, China ' Shihezi University, Yili – 835000, Xinjiang, China ' Shihezi University, Jiaozhuo – 454000, Henan, China

Abstract: To enhance the efficiency of replanting tray seedlings in greenhouse tomato cultivation, this paper proposes a robotic arm replanting path planning method based on an improved Q-learning algorithm. The objective is to reduce both path planning distance and computational time. Firstly, a dual-Q-table design aligns with the robotic arm's dual-tray movement logic, resolving mapping inconsistencies inherent in single-Q-table nodes. Secondly, an ant colony algorithm pheromone matrix is introduced to narrow the exploration scope by eliminating low-concentration cell holes. Subsequently, a multi-dimensional real-time feedback reward function is devised to evaluate each step, reinforcing global planning. Finally, a dynamic ε-greedy strategy enhances the algorithm's exploration and convergence capabilities. Under experimental conditions involving a 128-cell tray, this model reduced path planning distance by 4.37% compared to the SARSA algorithm, 1.85% compared to Q-learning, and 0.84% compared to the expected-SARSA model. Simulation results demonstrate that the enhanced Q-learning algorithm effectively optimises transplanting paths.

Keywords: tomato seedling trays; replanting pathway planning; reinforcement learning; pheromones; multidimensional real-time feedback rewards.

DOI: 10.1504/IJAHUC.2026.155489

International Journal of Ad Hoc and Ubiquitous Computing, 2026 Vol.52 No.3, pp.178 - 191

Received: 10 Nov 2025
Accepted: 30 Dec 2025
Published online: 03 Aug 2026 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article