Abstract
This research introduces a hybrid trajectory planning strategy for a Delta parallel tea-picking robot, optimizing for both operational time and motion smoothness (jerk). By combining a Delta mechanism with vision-based localization, the system utilizes Lamé curves to smooth transitions between vertical and horizontal movements, preventing acceleration spikes. An improved 15-stage harmonic jerk model, optimized via Particle Swarm Optimization (PSO), is applied to maximize efficiency while adhering to kinematic constraints. Experimental results demonstrate that this approach reduces picking time by up to 23% and significantly suppresses mechanical vibration, offering a robust solution for automated high-quality tea harvesting.
Introduction
High-quality tea harvesting is a time-sensitive process requiring extreme precision. While Delta parallel robots are favored for their high-speed capabilities, they face a significant mechanical challenge: "acceleration spikes" during the transition between lifting and horizontal translation. These spikes cause excessive jerk, leading to vibrations that degrade picking accuracy and reduce the machine's lifespan. This paper addresses these issues through a multi-layered trajectory optimization strategy.
System Architecture and Methods
The proposed system integrates high-speed robotics with advanced computer vision to identify and harvest tea buds in real-time.
Delta Mechanism & Vision
The hardware consists of a 3-DOF Delta parallel robot mounted on a crawler chassis. Dual cameras provide 3D localization of tea buds, which are then processed to generate picking coordinates.
Trajectory Optimization Strategy
The core innovation lies in the hybrid trajectory design:
- Lamé Curve Smoothing: Traditional "gate" trajectories have sharp corners that cause instantaneous acceleration jumps. The authors employ Lamé curves to create smooth transitions between the vertical (lifting/descending) and horizontal segments. By adjusting the curve parameter $m$, the path can be tuned to balance workspace efficiency and motion continuity.
- 15-Stage Harmonic Jerk Model: While standard models use 7 stages, this paper introduces a 15-stage model. This expansion allows the robot to maintain peak acceleration and jerk for longer durations without exceeding physical limits, effectively "flattening the peaks" of the motion profile to increase speed without increasing vibration.
- Adaptive PSO Optimization: A Particle Swarm Optimization algorithm determines the optimal time intervals for each segment. A resonance coefficient ($\alpha$) is used to tune the trade-off between raw speed and motion smoothness.
Key Results
The strategy was validated using spatial data from Laoshan tea gardens across varying bud densities:
- Efficiency: The hybrid method reduced total picking time by 14.70% to 23.13% compared to traditional line-by-line picking.
- Model Performance: The 15-stage model outperformed standard harmonic jerk models, shortening the picking cycle by an additional 2.78% to 4.66%.
- Vibration Suppression: Peak jerk was reduced from approximately 26 mm/s³ to 7.5 mm/s³, a massive improvement in mechanical stability.
Conclusion
By combining geometric path smoothing (Lamé curves) with time-domain jerk optimization, this research provides a comprehensive framework for high-performance agricultural robotics. Future work will focus on environmental robustness under varying outdoor lighting and integrating dynamic obstacle avoidance for complex field conditions.
Summary updated