แนวคิดหลัก
Automating labor-intensive horticultural tasks with the HortiBot system for selective harvesting of sweet peppers.
บทคัดย่อ
I. Introduction
Labor-intensive horticultural tasks require automation.
Challenges in robotic manipulation due to semi-structured workspaces and varying environmental conditions.
Need for gentle manipulation of non-rigid plant organs.
II. Related Work
Automation in horticulture focusing on fruit detection and localization.
Selective harvesting systems using specialized hardware.
Review of existing robotic solutions for selective harvesting.
III. Hardware Setup and System Overview
Dual-arm manipulation system with an articulated head for perception.
Use of xArm7 and Lite6 with sensors and end-effectors.
Workspace analysis, calibration, and system overview.
IV. Perception and World Modeling
Plant organ detection using multi-dataset approach.
3D mapping, shape completion, and pepper plant modeling.
Online fruit following and pose update during manipulation.
V. Dual-Arm Manipulation
Parameterized Motion Primitives (PMP) for fixed goal poses.
Online Trajectory Generation (OTG) with collision checking.
Adaptive manipulation based on sensor measurements.
VI. Results
Superior peduncle detection performance compared to existing methods.
Experimental setup for sweet pepper harvesting trials.
Evaluation of success rate, execution time, and system performance.
สถิติ
Submitted to International Conference on Intelligent Robots and Systems (IROS) 2024.