Chinese Researchers Simulate Harvesting Robot Dynamics for Smoother Agricultural Automation
In a bid to refine the precision and reliability of agricultural robotics, a team of engineers from Henan Polytechnic Institute has successfully conducted a high-fidelity dynamic simulation of a fruit-picking robot using advanced multibody dynamics software. The research, led by Shi Chen and Lei Lei, focuses on optimizing the mechanical performance of robotic harvesters through virtual prototyping and simulation—critical steps toward enabling autonomous, impact-free operation in delicate orchard environments.
As global agriculture faces mounting pressure to increase productivity while reducing labor dependency and environmental impact, robotic harvesting has emerged as a promising solution. Yet, despite rapid advancements in artificial intelligence and computer vision, the mechanical execution of harvesting tasks remains a significant engineering challenge. A robot must not only locate and identify ripe fruit but also maneuver its arm with precision, applying just enough force to pick without damaging the crop or the plant. This requires a deep understanding of the robot’s dynamic behavior under real-world motion profiles.
The study addresses this challenge by leveraging a combined CAD-simulation workflow, starting with the design of a six-degree-of-freedom harvesting robot in SolidWorks, a professional 3D mechanical design platform. The digital model, simplified to exclude minor fasteners such as screws and washers to enhance computational efficiency, was then imported into ADAMS (Automatic Dynamic Analysis of Mechanical Systems), a leading multibody dynamics simulation environment widely used in automotive, aerospace, and robotics industries.
ADAMS enables engineers to simulate the physical interactions between moving parts, including forces, torques, velocities, and accelerations, under various motion scenarios. Unlike pure kinematic analysis, which only examines motion without considering forces, dynamic simulation accounts for inertia, gravity, friction, and joint interactions—providing a far more realistic representation of how a machine will behave in operation.
The research team followed a structured simulation workflow, beginning with the reconstruction of mechanical constraints and material properties within ADAMS. Once the virtual assembly was complete, they defined joint types—revolute, cylindrical, spherical, and fixed—based on the intended motion of each component. For instance, the connection between the chassis and the main arm was modeled as a fixed joint, while the rotating platform and the large arm were linked with revolute joints to allow controlled angular movement.
A critical phase of the study involved path planning for the robot’s end effector—the mechanical hand responsible for grasping and detaching fruit. Inspired by typical harvesting motions, the researchers programmed a “door-shaped” trajectory: the arm rises vertically, moves horizontally toward the target, descends to pick the fruit, then reverses the path to return the produce to a collection bin. This motion pattern, common in vertical farming and orchard robotics, places specific demands on the robot’s joints, particularly in terms of torque generation and vibration control.
To simulate this complex motion with high accuracy, the team employed a fifth-order STEP function—a smooth, continuous mathematical function that avoids abrupt changes in velocity and acceleration. This choice is crucial in robotics, where sudden jerks can lead to mechanical stress, reduced lifespan, and inaccurate positioning. By using a fifth-order polynomial, the researchers ensured that both acceleration and jerk (the rate of change of acceleration) remained continuous, minimizing dynamic shocks during operation.
The simulation was run over a 3-second period with 500 time steps, providing a detailed temporal resolution of the robot’s behavior. Using ADAMS’ post-processing tools, the team analyzed the velocity and acceleration profiles of the end effector along the X, Y, and Z axes. The results were striking: all motion curves were smooth and continuous, with no discontinuities or abrupt spikes. This indicates that the robot’s joints moved in a coordinated and stable manner, free from sudden impacts or oscillations that could compromise performance or damage delicate fruit.
From an engineering standpoint, these findings validate the mechanical design of the robot. Smooth velocity and acceleration profiles suggest that the inertial forces, gravitational loads, and joint torques are well-balanced across the system. This balance is essential for ensuring that motors and gearboxes are neither overburdened nor underutilized, thereby optimizing energy efficiency and component longevity.
The study also delved into the underlying dynamics of the robot using Lagrangian mechanics, a powerful framework for modeling multi-link robotic systems. By formulating the system’s kinetic and potential energy, the researchers derived a set of dynamic equations that describe how joint torques relate to angular positions, velocities, and accelerations. While the full equations involve complex terms related to inertia matrices, Coriolis forces, centrifugal effects, and gravitational components, the simulation allowed for numerical evaluation without requiring manual solution of these equations.
One of the key insights from the analysis was the relative significance of different dynamic forces across the robot’s joints. For the larger, slower-moving joints—such as the base rotation (S joint) and the main arm swing (L joint)—Coriolis and centrifugal forces were found to be negligible due to their limited angular range and low speed. However, for the smaller, faster-moving joints near the wrist and forearm, these inertial effects became significant and could not be ignored in the torque calculations.
Similarly, the mass and inertia of the upper arm (U joint) and other heavy components played a dominant role in the overall dynamic load. This highlights the importance of accurate mass distribution and inertia modeling in simulation—small errors in these parameters can lead to large discrepancies in predicted torque requirements, potentially resulting in motor undersizing or excessive power consumption.
By identifying which joints contribute most to the dynamic load, the researchers were able to prioritize design improvements and component selection. For example, high-torque, low-backlash gearboxes may be justified for the U joint, while lighter, faster actuators could suffice for the wrist joints. This level of insight is difficult to achieve through physical prototyping alone, which is often time-consuming and expensive.
The successful simulation also confirmed that the virtual prototype had the correct number of degrees of freedom—zero, after all six joints were properly constrained. This verification step, performed using ADAMS’ built-in model-checking tool, ensured that the digital model was neither under-constrained (which would lead to unrealistic motion) nor over-constrained (which would prevent motion altogether). A well-constrained model is essential for reliable simulation results.
Beyond mechanical validation, the study has broader implications for the future of smart farming. As labor shortages continue to affect agricultural sectors worldwide, especially in fruit picking—a task that remains largely manual due to its complexity—automated solutions are becoming increasingly attractive. However, deploying robots in unstructured outdoor environments presents unique challenges, including variable lighting, uneven terrain, and unpredictable plant geometry.
While much of the current research focuses on perception—using cameras and AI to detect fruit—the mechanical execution side has received less attention. This study helps bridge that gap by demonstrating how simulation can be used to optimize the physical interaction between robot and plant. A robot that moves smoothly and predictably is not only more efficient but also safer for both crops and human workers.
Moreover, the use of virtual prototyping significantly reduces development time and cost. Instead of building multiple physical prototypes and testing them in the field, engineers can iterate rapidly in software, adjusting link lengths, joint types, and motion profiles to achieve optimal performance. This accelerates innovation and allows smaller institutions and startups to compete in the agricultural robotics space.
The work also underscores the importance of interdisciplinary collaboration. Designing a functional harvesting robot requires expertise in mechanical engineering, control systems, materials science, and software development. By integrating CAD, dynamics simulation, and control theory, the research team exemplifies how modern engineering problems are solved through a systems approach.
Looking ahead, the next steps could include incorporating more realistic environmental factors into the simulation, such as wind loads, fruit stiffness, and branch flexibility. Additionally, coupling the mechanical model with a control system simulation would allow for closed-loop testing of trajectory tracking and disturbance rejection—key capabilities for real-world deployment.
Another promising direction is the integration of machine learning to optimize motion planning. While the current study uses a predefined path, future robots could adapt their trajectories in real time based on sensor feedback. Simulating such adaptive behaviors in a virtual environment would be a logical next step.
The research also opens the door to energy efficiency analysis. By examining the power consumption of each joint over the harvesting cycle, engineers could identify opportunities for regenerative braking or energy recovery—particularly useful in battery-powered field robots.
In practical terms, the findings could inform the design of commercial harvesting robots for apples, pears, and other tree fruits. Companies such as FFRobotics, Tevel, and Agrobot are already developing such systems, but widespread adoption has been limited by cost, reliability, and performance. Studies like this one provide the foundational engineering data needed to build more robust and cost-effective machines.
Furthermore, the methodology is not limited to fruit harvesting. The same simulation techniques can be applied to other agricultural robots, including pruning systems, spraying platforms, and weeding machines. As the agricultural sector embraces digital transformation, tools like ADAMS will play an increasingly central role in the design and validation of intelligent farming equipment.
The success of this study also reflects broader trends in Chinese engineering research. Over the past decade, Chinese institutions have made significant investments in robotics, automation, and precision agriculture. This paper, supported by a key scientific research project from the Henan Provincial Department of Education, is part of that growing momentum.
While much of the global attention has focused on AI and software innovation, this research reminds us that hardware and mechanical design remain critical. No matter how intelligent a robot’s brain is, its effectiveness ultimately depends on the precision and reliability of its body.
In conclusion, the work by Shi Chen and Lei Lei demonstrates the power of virtual prototyping in advancing agricultural robotics. By simulating the dynamic behavior of a harvesting robot with high fidelity, they have provided valuable insights into joint stability, motion smoothness, and torque requirements. Their results confirm that the mechanical design is sound and capable of stable, impact-free operation—essential qualities for any robot entering the delicate world of fruit picking.
As the world seeks sustainable and efficient food production methods, studies like this one pave the way for smarter, more capable machines that can work alongside farmers to meet the challenges of the 21st century. The future of farming may well be automated—but it will be built on solid engineering principles, validated not just in the field, but in the virtual lab.
Shi Chen, Lei Lei, Henan Polytechnic Institute, Journal of Agricultural Mechanization Research, DOI: 10.13300/j.cnki.cn41-1195/s.2021.08.006