Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Towards Fully Automated Decision-Making Systems for Greenhouse Control: Challenges and Opportunities

Yongshuai Liu, Taeyeong Choi, Xin Liu

TL;DR

This survey analyzes policy-learning approaches for fully autonomous greenhouse control, framing farm management as an MDP/POMDP with safety and resource constraints. It covers methods from RL, Bayesian optimization, imitation learning, model-based RL, and Sim2Real, emphasizing data efficiency, safety, and human–AI collaboration. The authors share practical lessons from their second-place finish in the 3rd Autonomous Greenhouse Challenge, extracting design insights about data scarcity, cross-simulator generalization, and delayed rewards. The paper proposes a comprehensive set of opportunities—informative sampling, knowledge-guided learning, model learning, meta-learning, explainable AI, and multimodal ML—to guide future research and real-world deployments. Overall, it offers a structured, methodical view of challenges and actionable directions for deploying autonomous AI in farming systems.

Abstract

Machine learning has been successful in building control policies to drive a complex system to desired states in various applications (e.g. games, robotics, etc.). To be specific, a number of parameters of policy can be automatically optimized from the observations of environment to be able to generate a sequence of decisions leading to the best performance. In this survey paper, we particularly explore such policy-learning techniques for another unique, practical use-case scenario--farming, in which critical decisions (e.g., water supply, heating, etc.) must be made in a timely manner to minimize risks (e.g., damage to plants) while maximizing the revenue (e.g., healthy crops) in the end. We first provide a broad overview of latest studies on it to identify not only domain-specific challenges but opportunities with potential solutions, some of which are suggested as promising directions for future research. Also, we then introduce our successful approach to being ranked second among 46 teams at the ''3rd Autonomous Greenhouse Challenge'' to use this specific example to discuss the lessons learned about important considerations for design to create autonomous farm-management systems.

Towards Fully Automated Decision-Making Systems for Greenhouse Control: Challenges and Opportunities

TL;DR

This survey analyzes policy-learning approaches for fully autonomous greenhouse control, framing farm management as an MDP/POMDP with safety and resource constraints. It covers methods from RL, Bayesian optimization, imitation learning, model-based RL, and Sim2Real, emphasizing data efficiency, safety, and human–AI collaboration. The authors share practical lessons from their second-place finish in the 3rd Autonomous Greenhouse Challenge, extracting design insights about data scarcity, cross-simulator generalization, and delayed rewards. The paper proposes a comprehensive set of opportunities—informative sampling, knowledge-guided learning, model learning, meta-learning, explainable AI, and multimodal ML—to guide future research and real-world deployments. Overall, it offers a structured, methodical view of challenges and actionable directions for deploying autonomous AI in farming systems.

Abstract

Machine learning has been successful in building control policies to drive a complex system to desired states in various applications (e.g. games, robotics, etc.). To be specific, a number of parameters of policy can be automatically optimized from the observations of environment to be able to generate a sequence of decisions leading to the best performance. In this survey paper, we particularly explore such policy-learning techniques for another unique, practical use-case scenario--farming, in which critical decisions (e.g., water supply, heating, etc.) must be made in a timely manner to minimize risks (e.g., damage to plants) while maximizing the revenue (e.g., healthy crops) in the end. We first provide a broad overview of latest studies on it to identify not only domain-specific challenges but opportunities with potential solutions, some of which are suggested as promising directions for future research. Also, we then introduce our successful approach to being ranked second among 46 teams at the ''3rd Autonomous Greenhouse Challenge'' to use this specific example to discuss the lessons learned about important considerations for design to create autonomous farm-management systems.

Paper Structure

This paper contains 29 sections, 1 equation, 3 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Background
  3. Autonomous Greenhouse Control
  4. Lettuce Growth Simulator
  5. Farm Management as a Constrained MDP
  6. Practical Challenges
  7. Complexity of Agricultural Systems
  8. Resource Constraints & Safety Requirements
  9. Slow Growth Rates
  10. Interaction with Human Farmers
  11. Potential Opportunities
  12. Informative Sampling
  13. Bayesian Optimization
  14. Novelty-driven Search
  15. Knowledge-guided Learning
  16. ...and 14 more sections

Figures (3)

  • Figure 1: Illustration of greenhouse control system where a centralized controller autonomously utilize the actuators and human to maximize productivity based on various types of data from sensors.
  • Figure 2: Workflow with BO, in which P and S represent the net profit and the control setpoints, respectively.
  • Figure 3: Prediction examples in our MBRL method: \ref{['fig:pred_temp1']} and \ref{['fig:pred_temp2']} are the $1$- and $20$-step lookahead predictions of indoor temperature, respectively while \ref{['fig:pred_plant1']} and \ref{['fig:pred_plant2']} are those of fresh weight of lettuce.