Robotic Bin-Picking Solution for Translucent Bags | YK-Robotics

At a glance

Challenge

Manual handling of bagged materials can be inefficient and unsafe. Industries are looking to automate these tasks, but translucent, deformable bags — randomly piled, varying in weight, and often overlapping — push the limits of both 3D vision and robotic manipulation.

Solution

YK-Robotics built a complete robotic cell combining a 6-DOF robot arm, Zivid 2+ 3D camera, and proprietary perception and planning software to reliably detect, segment, and pick transparent bags of different shapes and weights.

Result

The deployed system achieves average cycle times under 15 seconds, dynamically optimizing performance across up to eight bins based on bin state and recipe requirements.

Coherent pixel-wise depth data from Zivid 2+ was critical for training our segmentation model and reliably separating translucent bags in situations where RGB data alone was insufficient.”

Joan Ortega Alcaide, CEO at YK-Robotics

01 Story

YK-Robotics has developed a robotic bin-picking solution designed for the automated handling of bagged materials. The application enables an industrial robot to detect, segment, and pick translucent bags from bins and place them onto a conveyor belt according to a predefined, recipe-based workflow. The solution targets industries where bulk materials are commonly handled in bags, including chemicals, materials processing, recycling, food ingredients, and general manufacturing.

The project originated from customer demand. Manual handling of bagged materials was inefficient, ergonomically challenging, and unsafe, while existing automation solutions struggled with translucent packaging, highly variable contents, and random piling. These limitations exposed a clear gap in the market for a robust 3D-vision-based system capable of handling deformable, visually complex objects.

02 Challenges

The application posed several significant technical challenges across perception, robotics, and software. From a vision perspective, the system needed to reliably perceive translucent, deformable bags whose contents vary widely in shape, texture, color, and size. The bags weigh between 1.5 kg and 12 kg and are randomly piled in bins of different shapes and materials, often overlapping and partially occluded. These characteristics make both visual separation and mechanical handling difficult.

On the robotics side, the robot had to safely manipulate heavy yet delicate objects while maintaining short cycle times. This required careful balance between speed and gentle handling, with robot motion dynamics optimized online on a cycle-by-cycle basis.

Software complexity added another layer of difficulty. The system had to perform robust segmentation, select safe and reliable pick points, and continuously optimize pick order and trajectories under high uncertainty. Achieving this while maintaining high throughput placed strict requirements on depth accuracy, dimensional trueness, spatial resolution, and stable color information from the vision system.

03 Solution

Development was primarily carried out in-house at YK-Robotics, with iterative testing and validation on real customer installations to ensure robustness under realistic conditions. L.C.E. Engineering and Robotic S.r.l. have designed the gripper, built and mounted the entire cell while providing the space and support for the several months of tests done. The solution is delivered as a complete robotic cell, composed of:

A 6-DOF industrial robotic arm
A Zivid 2+ L110 3D camera mounted on the robot arm
Up to eight interchangeable bins with varying materials, shapes, and sizes
End-of-arm tooling designed for bag handling
A conveyor belt
An orchestrator system defining recipes and phase constraints
YK-Robotics’ in-house perception, planning, and control software platform

YK-Robotics chose the Zivid 2+ camera for several key features. Getting coherent, pixel-wise data from the 3D camera was essential. Depth information from the camera was used to train segmentation models capable of distinguishing individual bags in scenarios where RGB data alone was insufficient or unreliable. Having ordered, aligned 3D, 2D, color, and surface information significantly simplified downstream perception and planning.

The Zivid SDK integrated smoothly into YK-Robotics’ software stack, allowing the team to focus on higher-level perception and motion planning rather than low-level camera control. Technical support played a decisive role early in the project, helping convince the end customer that the problem was solvable and building confidence to initiate this technically demanding application.

What sets the solution apart is its ability to combine reliable perception of translucent, deformable bags with multi-bin handling, recipe-based orchestration, and dynamic optimization of pick order and robot trajectories under strict safety and cycle-time constraints.

04 Result

The overall project took several months, with approximately nine months dedicated specifically to 3D machine vision development, focusing on segmentation and pick-point evaluation for translucent bags.

The deployed system can operate with up to eight bins simultaneously and dynamically optimizes cycle time based on bin state and recipe requirements. Average cycle time is under 15 seconds, with exact performance depending on the specific customer configuration.

The end customer is satisfied with the performance achieved and is already considering deploying the same solution across additional production lines, confirming both technical success and commercial viability.

About YK-Robotics

YK-Robotics is an industrial robotics company headquartered in Florence, Italy. The company develops and deploys advanced robotic automation solutions for manufacturers and system integrators, covering a wide range of applications including bin picking, assembly, palletizing, dispensing, and packaging. At the core of their offer is the YK-Platform, a proprietary software environment that covers the full automation lifecycle — from digital twin design and simulation to on-site deployment and custom HMI development.

→ yk-robotics.com