Cucumber Picking Robot Uses Fuzzy Logic for Precision Harvest
In the rolling fields of Xinjiang, where sunlight bathes endless rows of greenhouses, a quiet revolution is taking root—one that could redefine the future of agricultural labor. As rural populations dwindle and the farming workforce ages, automation is no longer a luxury but a necessity. Against this backdrop, a team of researchers from Xinjiang Shihezi Vocational and Technical College has developed a novel cucumber harvesting robot that combines streamlined mechanics with intelligent control systems to deliver high accuracy and reliability in real-world conditions.
Led by Zhang Wei, Mu Huiqun, and Guo Jian, the research team has introduced an autonomous harvesting system that addresses long-standing challenges in agricultural robotics: bulkiness, low success rates, high costs, and imprecise motion control. Their solution, detailed in a recent publication in Agricultural Mechanization Research, integrates a three-degree-of-freedom robotic arm with a track-based mobile platform and leverages fuzzy logic algorithms to enhance trajectory precision during harvesting operations.
The study, titled “The Application of Fuzzy Algorithm in Trajectory Control for Cucumber Picking Robot,” presents a comprehensive approach to robotic harvesting that begins with mechanical design and extends into advanced control theory. Unlike many existing agricultural robots that rely on complex, multi-jointed arms and high-cost sensors, this system prioritizes simplicity, durability, and scalability—qualities essential for widespread adoption in commercial farming.
At the heart of the robot is its articulated harvesting arm, composed of three primary joints: a waist joint for horizontal rotation, an upper arm, and a lower arm. These joints work in concert to position the end-effector—equipped with gripping and cutting mechanisms—precisely at the base of each cucumber. The entire system is mounted on a dual-track mobile base, allowing it to navigate narrow rows in greenhouse environments with stability and minimal soil compaction.
One of the most critical aspects of robotic harvesting is motion planning. Traditional robotic arms often use high-order polynomial functions to define joint trajectories, which can lead to jerky movements, mechanical stress, and positioning errors—especially when starting or stopping. To overcome this, the Shihezi team adopted a cycloid function for trajectory planning. This mathematical approach ensures that both velocity and acceleration are zero at the beginning and end of each motion segment, effectively eliminating mechanical shock and vibration.
The cycloid-based motion profile allows the robotic arm to move smoothly from one position to another, even at higher speeds. This not only reduces wear on the actuators but also increases the overall cycle time for harvesting. In practical terms, smoother motion translates into more consistent positioning, which directly impacts the success rate of each pick.
However, even with optimal trajectory planning, real-world factors such as friction, gear backlash, and load variations introduce uncertainties that can degrade performance. These disturbances are particularly pronounced in lightweight robotic systems where small deviations can accumulate across multiple joints. To compensate for these nonlinearities, the team implemented a fuzzy logic-based control system designed to approximate and counteract the uncertain friction terms in the robot’s dynamic model.
Fuzzy logic, a form of artificial intelligence that deals with reasoning that is approximate rather than fixed and exact, is especially well-suited for systems where precise mathematical models are difficult to derive. In this application, the fuzzy controller continuously monitors the deviation between the desired and actual joint positions, then adjusts the control input to minimize error. Unlike traditional PID controllers, which rely on fixed gain parameters, fuzzy systems adapt their response based on linguistic rules—such as “if the error is large, apply a strong correction”—making them more robust in unpredictable environments.
The integration of fuzzy control into the robot’s dynamics represents a significant advancement in agricultural robotics. By modeling the friction and other unmodeled dynamics as uncertain terms, the controller effectively “learns” how to compensate for them in real time. This adaptive capability ensures that the end-effector remains on target, even as external conditions change—such as variations in cucumber weight, stem stiffness, or actuator temperature.
To validate the effectiveness of their design, the researchers conducted a series of laboratory and field tests. In the first phase, they evaluated the robot’s positioning accuracy by commanding it to move to ten predefined target coordinates and measuring the deviation between the intended and actual end-effector positions. The results showed that the maximum error was 6.1 mm in the y-direction, with standard deviations below 1 mm in all axes. Given that the average cucumber is over 200 mm in length, this level of precision is more than sufficient for successful harvesting.
More importantly, the error distribution was consistent and repeatable, indicating that the system’s control algorithms were effectively managing the dominant sources of inaccuracy. The z-axis, which corresponds to vertical positioning, exhibited the smallest deviation, suggesting that the robot’s kinematic model and gravity compensation were well-tuned. The x- and y-axes, which involve more complex joint interactions, showed slightly higher variability but still remained within acceptable limits.
With the mechanical and control systems proven in controlled conditions, the team moved to real-world testing. Over multiple sessions in active cucumber greenhouses, the robot performed 80 harvesting attempts across four separate test groups, with 20 picks per group. The overall success rate reached 89.95%, with individual test runs ranging from 86.6% to 93.3%. Each successful harvest took an average of 16.34 seconds, with the fastest runs completing in just 14.6 seconds.
These results place the Shihezi robot among the most efficient cucumber harvesters reported in recent literature. While some experimental systems have achieved higher speeds, they often do so at the expense of reliability or require highly structured environments. In contrast, this robot operates effectively in typical greenhouse conditions, where lighting, plant spacing, and fruit orientation vary significantly.
The success of the system can be attributed to the synergy between its mechanical design and intelligent control. The three-degree-of-freedom arm, while simpler than six-axis industrial robots, is well-matched to the task at hand. Cucumbers grow in relatively predictable orientations, and the robot’s workspace is sufficient to reach fruits within a standard planting layout. By avoiding unnecessary complexity, the team reduced both cost and points of failure.
Moreover, the use of off-the-shelf servo motors and a modular mechanical structure makes the robot easier to maintain and repair—a crucial consideration for farmers who may lack access to specialized technicians. The track-based mobility system further enhances practicality, allowing the robot to operate on uneven or soft terrain without getting stuck.
From a control theory perspective, the implementation of fuzzy logic represents a pragmatic solution to a persistent problem in robotics: how to maintain accuracy in the presence of unmodeled dynamics. While more advanced techniques such as neural networks or model predictive control could potentially offer higher performance, they also require greater computational resources and extensive training data—factors that hinder deployment in cost-sensitive agricultural applications.
The fuzzy system used in this robot strikes a balance between sophistication and simplicity. It does not attempt to model every physical parameter of the arm but instead focuses on correcting the most significant sources of error. This targeted approach ensures that the control system remains efficient and responsive, even on modest hardware.
Another notable aspect of the research is its emphasis on real-world validation. Many robotics studies remain confined to laboratory environments, where conditions are tightly controlled and variables are minimized. In contrast, the Shihezi team tested their robot in actual farming conditions, exposing it to the variability and unpredictability that define real agriculture. The fact that the system maintained a high success rate under these conditions speaks to its robustness and practical viability.
The implications of this work extend beyond cucumber harvesting. The principles of simplified mechanics, smooth trajectory planning, and adaptive control could be applied to other crops such as tomatoes, peppers, or strawberries—each of which presents unique challenges but shares the need for gentle, accurate manipulation. As labor shortages continue to affect global agriculture, systems like this one could play a pivotal role in sustaining food production.
Looking ahead, the researchers suggest several avenues for improvement. Integrating machine vision systems could allow the robot to autonomously detect and classify cucumbers, reducing reliance on external guidance. Enhancing the end-effector design could improve grip stability and reduce the risk of fruit damage. Additionally, deploying multiple robots in a coordinated fleet could increase coverage and efficiency in large-scale operations.
The economic case for such robots is becoming increasingly compelling. While the initial investment may be significant, the long-term savings in labor costs, combined with the ability to harvest at optimal times, can yield substantial returns. Furthermore, automated systems can operate consistently without fatigue, ensuring that crops are harvested at peak quality.
Environmental benefits may also emerge. Precise harvesting reduces waste by minimizing damage to plants and unripe fruits. Efficient path planning and electric-powered mobility contribute to lower emissions compared to traditional machinery. As sustainability becomes a central concern in agriculture, intelligent robots offer a path toward more responsible farming practices.
The work of Zhang Wei, Mu Huiqun, and Guo Jian exemplifies how innovation in agricultural technology does not always require cutting-edge components or massive budgets. Sometimes, the most impactful advances come from thoughtful integration of existing technologies, guided by a deep understanding of both engineering principles and real-world needs.
Their cucumber harvesting robot is not just a machine; it is a response to a growing crisis in rural labor. It is a tool that empowers farmers to maintain productivity in the face of demographic shifts. It is a demonstration that automation, when designed with practicality and resilience in mind, can serve not only industry but also society.
As the global population continues to rise and arable land becomes scarcer, the challenge of feeding the world will only intensify. Solutions like this one—rooted in engineering ingenuity and field-tested reliability—offer a glimpse of a future where technology and agriculture grow together, hand in hand.
In an era where artificial intelligence often conjures images of distant, abstract systems, this robot stands as a reminder that the most meaningful applications of AI are those that touch the earth, that harvest the food we eat, and that support the people who grow it. The cucumber picking robot from Xinjiang Shihezi Vocational and Technical College is not just harvesting vegetables—it is cultivating the future of farming.
Zhang Wei, Mu Huiqun, Guo Jian, Xinjiang Shihezi Vocational and Technical College, Agricultural Mechanization Research, DOI: 10.3969/j.issn.1003-188X.2021.03.034