--- title: "China's picking robots harvest 5,000 apples per charge as orchards face a labour gap" date: 2026-10-05 category: Humanoids & Robotics site: NeuroAI canonical: https://neuroai.site/a/na-robot-agri-picking language: en --- # China's picking robots harvest 5,000 apples per charge as orchards face a labour gap > From apple arms to tomato grippers, Chinese picking robots now work real orchards and greenhouses, with dual-arm models picking 5,000 apples per charge. In a terraced apple orchard in Gansu, a pair of robotic arms reaches into the canopy, twists a ripe fruit free, and drops it gently into a bin — while the farmer who once did this by hand watches from the shade. Across China, the "who will pick the fruit" problem is meeting a mechanical answer. The 2026 Central No. 1 Document (中央一号文件) listed "agricultural robots" (农业机器人) as a category for the first time, a signal that picking automation has moved from research curiosity to national priority. ## Apples: the dual-arm worker Beijing Academy of Agriculture's Intelligent Equipment Centre built a dual-arm orchard picker now demonstrated in Tianshui's Huaniu apple base. Charged for 2 hours, it runs 8 hours and can pick more than 5,000 apples — roughly double a human's pace. A companion autonomous transport robot carries 400–600 jin (≈ 200–300 kg) per trip on RTK navigation, handling 40-degree slopes and mud. ## Pears, tomatoes and the "specialised team" No single robot fits every crop, so a crop-specific "team" is emerging: - Pear: a three-arm picker in Danyang mimics the farmer's "lift and pull" so the stem breaks cleanly; ~300 jin (≈ 150 kg) per hour, ~2.5% damage. - Tomato: a multi-arm greenhouse robot in Jiangsu trims a ripe cluster in about 5 seconds using AI vision; the institute targets under 4 seconds and 24-hour runs. - Apple (multi-arm): Nanjing institute's version worked 8+ months in fields during 2025, ~1,200 apples/hour, under 5% damage. - Asparagus: a high-speed cutter harvests 2,500 stalks/hour, 24/7. ## Why now Two forces converge. Labour: farming ages, and in tomato production picking is nearly half the human cost (50–60% of total). Fruit like apples and pears ripen in a short window, so speed decides market timing. Tech: AI vision now tells fruit from leaf and stem, judges ripeness, and decides whether to grip — then a flexible end-effector avoids bruising. Already, smart-orchard pilots pair picking with digital management: soil sensors cut irrigation water from 40 to 20 m³ per mu, and micro-spray systems shield fruit during heatwaves. Robots are one layer of a "digital brain" orchard, not a standalone fix. ## From lab to field Field work is harder than the bench. Lighting shifts, leaves occlude fruit, and green apples hide against green leaves. Nanjing institute researchers told Xinhua that getting a robot to pick cleanly for even one or two minutes in the wild was a real breakthrough, proved only after heavy in-field data correction. The 2025 acceptance of Jiangsu's industry project demanded hard deployment data — how many units delivered, productivity, failure rate — over papers and patents. ## The cost question Today a multi-arm picker costs about 200,000 yuan (≈ US$28,000 / HK$220,000); the transport robot about 20,000 yuan (≈ US$2,800 / HK$22,000). As arm and vision modules scale, makers expect costs to fall toward the "call a service team" model — like combine harvesters, a crew brings robots to your orchard for a fee. ## A global problem, a Chinese toolkit The labour squeeze behind these machines is not unique to China. Japan, South Korea, parts of Europe and North America all run farms with fewer young hands and rising wage bills, and fruit picking is among the hardest jobs to fill. What China contributes is a manufacturing base that can drive pickers and transport robots down to a price point small farms can share rather than own outright. Early export enquiries and pilot interest from Belt-and-Road agricultural projects suggest the dual-arm and autonomous-transport designs trialled in Gansu and Jiangsu could travel abroad with little redesign. Crucially, the service-model path matters as much as the hardware: as with combine harvesters, a shared crew can bring robots to an orchard for a fee, so a 200,000-yuan machine becomes accessible without a capital purchase. The real test will be multi-season reliability and local service networks, not a single harvest-festival demo — but the category has clearly crossed from lab curiosity to a deployable toolkit, and the 2026 Central No. 1 Document (中央一号文件) makes that national intent explicit. ## What readers can do now - Orchard owners should trial a dual-arm picker on one high-value crop (apples, pears) where labour cost is steepest before buying. - Ag-tech buyers should demand field-hours and damage-rate data, not demo clips; Jiangsu's 2025 acceptance required real in-orchard numbers, not papers. - Policymakers and co-ops can pool robots as a shared service so small farms access the 200,000-yuan machines without owning them. ## Honest limitations Performance figures come from CCTV, Xinhua and regional reports of institute trials and harvest-festival demos; independent, multi-season yield comparisons are limited. Damage rates (2.5–5%) are vendor/institute-reported under specific conditions. "5,000 apples per charge" is a demonstrated run, not a certified season average. This article covers fruit/vegetable picking only; it excludes weeding, spraying and planting robots, and it does not assess provincial subsidy details. --- Published by NeuroAI (https://neuroai.site/) — https://neuroai.site/a/na-robot-agri-picking Free to quote with attribution and a link to the original.