Improving greenhouse fruit-production control by integrating reinforcement learning into short-horizon model predictive control

Greenhouse fruit-production control aims to maximize the economic performance (fruit revenue minus operating costs) while operating within system constraints under external weather disturbances. Control methods need to balance the delayed economic benefit of fruit yield with current operating costs. For such problems, model predictive control (MPC) can handle constraints under future weather disturbances, but can become computationally demanding when using sufficiently long prediction horizons for (relatively large) nonlinear greenhouse fruit production models. In contrast, reinforcement learning (RL) can learn control policies before deployment while considering economic performance beyond MPC’s prediction horizon. However, desired indoor climate ranges are typically incorporated through penalty terms, and performance may degrade under unseen weather trajectories. This work proposes trajectory-selection RL–MPC , a framework that incorporates the economic benefit of fruit yield beyond the MPC prediction horizon into a finite-horizon optimization problem with a relatively short horizon (one hour). The framework uses an RL rollout trajectory to define a terminal region constraint and terminal cost. Next, a nonlinear MPC solves a short-horizon optimization problem with these terminal ingredients to find a local optimum. Finally, the framework selects and executes the first input from the trajectory with the better objective value, either from the MPC-predicted or the RL rollout trajectory. The method is applied to GreenLight, a large-scale nonlinear greenhouse tomato production model that exhibits stiff dynamics. The simulation results show that trajectory-selection RL–MPC with a one-hour prediction horizon matches the closed-loop performance of a high-performing guiding policy while improving over standalone MPC with the same horizon. On 25 weather trajectories held out during policy training, trajectory-selection RL–MPC achieved ∼ 52 % higher closed-loop performance than standalone RL and ∼ 80 % higher performance than MPC with the same horizon length.