Python · C/C++ · Embedded Systems · Force Control · Optimal Control

Multi-Arm Autonomous Apple Picking Robot

Overview

Built from the ground up, a fully autonomous bilateral apple picking system for dwarf spindle apple trees in narrow orchards. The system integrates 6 gantry-mounted robotic arms with 48 drive units under real-time control, apple ripeness detection algorithms, and force-feedback optimized grasping control. The achievement was featured on CCTV and widely reported.

My Role

Control System Technical Lead. Responsible for the complete master-slave control system architecture, multi-node distributed control design, and end-effector force-feedback optimized grasping algorithms.

Key Decisions

  • With up to 48 synchronized joint control nodes, time synchronization and real-time communication were critical — designed a distributed control scheme that guarantees deterministic timing.
  • For the apple grasping end-effector, selected soft-finger grippers with precision force control to minimize fruit damage after thorough evaluation.
  • Built the distributed control system on Nvidia Jetson Xavier with ROS, implementing a hierarchical master-slave architecture (Jetson-STM32-Motors) to separate perception/decision-making from real-time execution — a "fast-slow decoupling" pattern.