Benchmarking Controllers for Low-Cost Agricultural SCARA Manipulators.

Area of Science:

• Agricultural Engineering
• Robotics
• Control Systems

Background:

• Increasing global demand necessitates enhanced agricultural productivity with reduced resource input.
• Labor shortages during critical periods like harvest highlight the need for agricultural automation.
• High costs of commercial robotic manipulators limit adoption, driving research into low-cost alternatives.

Purpose of the Study:

• To develop and evaluate control strategies for low-cost SCARA manipulators in agriculture.
• To address challenges in low-cost manipulators, including inaccuracy, limited sensor feedback, and dynamic uncertainties.
• To benchmark the performance of Sliding Mode Control (SMC), Reinforcement Learning (RL), and a novel Proportional-Integral feedforward (PIFF) controller.

Main Methods:

• Implementation of three distinct control strategies: SMC, RL, and PIFF.
• Benchmarking controller performance based on response time, movement smoothness, and robustness to dynamic changes.
• Focus on a low-cost SCARA manipulator tailored for agricultural applications.

Main Results:

• SMC achieved the fastest response time but exhibited joint movement jitter.
• RL controller demonstrated sudden breaks and overshoot issues when reaching setpoints.
• PIFF controller offered the smoothest reference tracking but showed increased susceptibility to system dynamics variations.

Conclusions:

• Each controller presents trade-offs in performance for low-cost agricultural manipulators.
• SMC is suitable when speed is critical, provided jitter can be managed.
• PIFF offers smoother operation, ideal for applications sensitive to jerky movements, but requires careful tuning for stability.