UC Davis

Harvest-aid platforms are used in modern orchards to improve manual harvesting efficiency, safety, and ergonomics. Typically, workers stand at pre-set heights on a platform's multi-level deck, and each worker harvests fruits inside a canopy zone that is defined by the lowest and highest reach of the worker's arms. However, fruit distributions are non-uniform, and worker picking speeds vary, thus generating a mismatch between labor demand (incoming fruit rates) and labor supply (fruit picking rates) in each zone; this mismatch limits platform-based harvesting efficiencies. To alleviate this problem, we transformed a conventional harvesting platform into a collaborative robot (co-robot) platform. As the co-robotic platform travels forward, it estimates the incoming fruit distribution using a vision system, it measures each worker's picking speed using instrumented picking bags, and controls the heights of hydraulic lifts that move workers up and down. The model-based control algorithm maximizes the machine's harvesting speed by changing the height at which each worker harvests as a response to incoming fruit load because it matches fruit-picking labor supply and demand. Simulation experiments with pre-recorded fruit distribution data validated the approach and provided efficiency gains under various conditions. Apple-harvesting experiments were also performed in a commercial orchard, where 2307 kg of apples were picked: 1045 kg in variable-height zone harvesting mode, and 1262 kg in fixed zone harvesting mode, with workers at fixed heights that were set by the grower. Variable-height zone harvesting mode throughput was 327.6 kg/h vs. 298.8 kg/h for fixed zone harvesting mode at human-controlled platform moving speed, resulting in an improvement of 9.5%.