Port Lock Station - PLS System

The PLS System developed by DEEPLOCK uses AI intelligent algorithms to achieve accurate 3D corner positioning and 2D twist-lock type recognition, and then use robot fixture to remove and install twist-lock precisely. It has the following advantages:

Core Advantages

1. All Day Operation

The machine is fully enclosed and the camera is not affected by sunlight or rain.

2. Flexible Layout

PLS has three types of Auto Lock Machine: Type T, Type L and Type G. Type T is for middle twist-lock and L/G is for side ones. They are all independent unit and can be combined freely and flexibly for layout based on the actual situation of port.

3. Convenient Maintenance

Standard parts use global big brands like Siemens, Festo, etc. maintenance door. Devices, wiring and tubes are clearly identified.

4. Intelligence

AI Deep learning, accurately identifying twist-lock coordinates and types. Edge computing greatly reduce algorithm time.

5. Adapt to Diversity

The self-adaptive fixture V2.5 can solve over 95% of twist-locks which are classified into 25 categories by DEEPLOCK of over 300 types worldwide.

6. Safety

Machine safety design fully meet CE standard.

Key Components

AI Vision + Self-adaptive Fixture Robot

Six axis robot with self-adaptive fixture and AI vision camera, recognizes twist-lock coordinates and types, then remove and install automatically. Repeatability positioning accuracy is ± 0.05mm.

Twist-lock Storage Rack

It has a maximum storage capacity of over 80 pallets and 330 twist-locks which can be customized based on customer's requirement. The twist-lock is stably held in the pallet without shaking

Transfer Rack Module

Mobile transfer rack is used for twist-lock transferring during multi ships and multi berths operations.

AI Technology Application

The Port Lock Station of DEEPLOCK focuses on AI training model in container corner positioning and twist-lock type recognition. With a multimodal perception system and AI algorithm model, the test result at berth 4 of Longtan Port in Nanjing shows that the 3D camera positioning tolerance is ≤ 0.4mm, the 2D camera recognition accuracy reached 99.7%, a single container's operation efficiency is promoted 45%. The agile deployment with AI provides a practical paradigm for empowering the intelligent upgrade of port equipment

Fixture Robot Dynamic 3D Camera

The 3D AI camera accurately obtains the three-dimensional coordinates (XYZ) and attitude angles (WPR) of the lock hole center, providing a docking reference for the fixture robot.

Technical Difficulties:

  • Scenario interference:The high humidity of port leads to rusting and dust accumulation on the surface of corner, obstructing feature points
  • Attitude deviation:The tilting of truck parking (± 15°) causes changes in the spatial attitude of corner, and traditional template matching fails
  • Real time requirement:The positioning calculation needs to be done within 200ms to match fixture robot motion cycle.

2D Camera

Normal types of twist-lock at port include single braid lock, double braid lock, automatic lock (ISO 17712), and automatic cone lock, requiring 2D camera to accurately identify the type and output characteristic parameters (such as the diameter of the automatic lock cylinder)

Technical Difficulties:

  • Small difference in type:the automatic lock is similar to the automatic cone shape, only the lock cylinder structure is different, and the traditional algorithm is easily confused
  • Obstruction problem:Incomplete feature extraction caused by oil stains on the lock surface and container shadows
  • Real time requirement:Recognition must be done within 100ms to avoid affecting the job cycle.

Technical Index Comparison

Based on 2000 Container Samples

Index TypeTraditional Vision SolutionAI Vision SolutionImprovement
3D Corner Positioning Tolerance3mm0.4mm73.3%↓
2D Lock Recognition Accuracy88%99.7%13.3%↑
Robot Positioning Failure5%0.3%94%↓
Single Container Operation Time65s36s45%↑
Manual Intervention Rate15%0.5%96.7%↓