China's Coconut Tree Group Wants a Robot That Peels 360 Coconuts Per Hour: Welcome to the Real World of Factory Automation

The Tender

China's Coconut Tree Group (椰树集团) — famous for its long-running coconut juice brand and equally famous advertising style — published a technical tender seeking an automated coconut-peeling system with these requirements:

• Throughput: Minimum 360 coconuts per hour (6 per minute)
• Damage rate: Less than 1% — the edible meat must remain intact
• Output: Completely shell-free coconut meat ready for processing

It sounds simple. It is not.

Why Coconuts Are a Robotics Nightmare

The Physics

Each coconut is unique:

• Irregular shape — No two coconuts have the same geometry
• Variable size — Diameter ranges from 15-25cm
• Hard shell — Requires significant force to crack
• Fragile meat — Too much force destroys the edible portion
• Fibrous husk — Must be fully removed without damaging what's inside
• Water inside — The liquid sloshes, changing weight distribution

The Industrial Context

In a factory processing thousands of coconuts daily:

• Speed — 6 per minute per station means minimal time for sensing and adaptation
• Consistency — Every coconut must meet quality standards
• Cost — The machine must be cheaper than human labor
• Maintenance — Must operate in hot, humid, corrosive environments (coconut water is acidic)
• Safety — Industrial cutting tools operating at speed need fail-safes

Why This Is Hard for AI/Robotics

The fundamental challenge is manipulation of irregular, deformable objects — arguably the hardest problem in robotics:

1. No fixed reference frame — Each coconut requires real-time 3D scanning and adaptive gripping
2. Deformable objects — The coconut meat can shift, compress, or tear unpredictably
3. Tight tolerances — <1% damage means <4 failures per 400 coconuts
4. High throughput — Leaves milliseconds for each decision
5. No training data — There isn't a large dataset of coconut-peeling robot demonstrations

The Current State of the Art

What Exists

• Simple automated coconut cracking machines (crude, high waste, not suitable for premium processing)
• CNC-based approaches (too slow, too expensive, too rigid)
• Water jet cutting (precise but slow and expensive)
• Manual labor (current standard — fast, flexible, but inconsistent)

What's Missing

A system that combines:

• Adaptive sensing (vision + force feedback)
• Deformable object manipulation
• Industrial speed and reliability
• Affordable cost

This intersection doesn't exist commercially yet.

The Bigger Story

This tender is a microcosm of where AI and robotics actually stand in 2026:

The coconut problem sits squarely in the "struggles with" column. And it's not unique — similar challenges exist in meat processing, fruit harvesting, textile handling, and countless other industries.

Why It Matters

If someone solves this problem, the implications extend far beyond coconuts:

• Food processing — Automated handling of irregular agricultural products
• Agriculture — Robot harvesting of soft fruits and vegetables
• Manufacturing — Flexible assembly of non-standard components
• Healthcare — Gentle manipulation of biological tissues

The company that builds a general-purpose deformable-object manipulation system will unlock trillions in labor cost savings across industries.

Coconut Tree Group isn't just looking for a machine. They've accidentally defined one of the hardest unsolved problems in robotics.

Source: Zhihu Discussion