In industrial robots, the end-effector, the tool that directly interacts with the workpiece, is crucial for determining the quality of manufacturing tasks such as grinding, polishing, and deburring. However, traditional industrial robot systems mostly focus on the motion control of the robot body, even though the control performance of the end-effector often becomes the key bottleneck restricting processing quality. Therefore, talent in Intelligent Manufacturing Engineering programs must be cultivated to master high-precision motion control and constant force control methods for industrial robot end-effectors. To bridge this gap and considerably enhance the practical skills of students, this paper designs and develops an experimental platform and a corresponding teaching method centered on the motion control system of an industrial robot end-effector. The core objective is to enable students to systematically understand and implement motion control and constant force control techniques, with particular emphasis on force–position hybrid control in realistic grinding scenarios.

A dedicated experimental platform was developed, comprising a six-degree-of-freedom serial industrial robot, a 2R1T end-effector driven by voice coil motors, a Beckhoff industrial PC, ELMO servo drives, incremental encoders, and a six-axis force sensor. TwinCAT3 software was used to implement the control system, which comprised four functional modules: motion control, logic operation control, adaptive constant force variable admittance control, and a human–machine interface. A hierarchical experimental teaching method incorporating three progressively integrated experiments—single-axis motion control, multi-axis coordinated motion control, and constant force grinding control, the core experiment—was proposed. In the constant force experiment, students implemented force–position hybrid control and compared the performance of a PID-optimized admittance controller with that of a neural network adaptive constant force variable admittance controller based on an RBF network. Students performed the grinding task with a target contact force of 15 N, guided through the complete engineering cycle of parameter initialization, system activation, experimental execution, data acquisition, and performance evaluation.

Through the single-axis motion control experiment, students mastered three positioning modes and developed precision positioning skills. The multi-axis coordinated motion control experiment enabled students to achieve smooth trajectory tracking and synchronous multi-axis movement. In the constant force grinding control experiment, students implemented force–position hybrid control and compared two control strategies. The neural network adaptive controller achieved steady-state force fluctuation within ±1.5 N and an average force error of 0.16 N, considerably outperforming the PID-optimized admittance controller, whose fluctuation range and average error were ±3.2 N and 0.31 N, respectively, under identical conditions. These results clearly illustrate the superior adaptability and disturbance rejection capability of the neural network adaptive constant force variable admittance controller. Throughout the process, students gained hands-on experience in parameter tuning, real-time data monitoring, system debugging, and quantitative performance evaluation.

The developed experimental platform and proposed teaching method effectively guide students from foundational motion control skills to a comprehensive understanding of force–position hybrid control. By completing the full engineering cycle of parameter setting, experimentation, data analysis, and iterative optimization, students develop practical abilities in system debugging, performance evaluation, and controller comparison. The teaching method provides a replicable and effective approach for cultivating high-quality talent in intelligent manufacturing.