Mujin Solves Factory Automation Challenges with the Zivid 3 3D Vision Camera

01 Story

Based in Tokyo, Japan, Mujin is a global leader in robotics, addressing some of the most complex challenges in factory automation. Since introducing the world’s first teachless robot controller, MujinController, the company has enabled robots to handle variability and make real-time decisions without manual programming.

The collaboration between Mujin and Zivid began in 2019, when Zivid engaged with Mujin to gain deeper insight into customer needs within manufacturing and warehousing environments in Japan. Since then, Mujin’s continuous field feedback has played a critical role in shaping the development of Zivid’s next-generation 3D cameras.

In 2025, Zivid introduced the Zivid 3, designed specifically for long-distance robotic applications of up to 4.5 meters and beyond, including depalletization, large bin picking, and piece picking. Mujin was among the first companies to test the Zivid 3 ahead of its launch and became one of its earliest customers.

02 Challenges

True to its reputation, Mujin works with some of the most demanding bin-picking and depalletizing applications, often involving highly reflective, shiny industrial parts. One of the key challenges was capturing complete pallets or large bins in a single scan from a distance of 3–4 meters, while still ensuring accurate and reliable 3D data.

Capturing reflective parts for bin-picking is inherently difficult, and doing so from several meters away adds further complexity. Many long-distance 3D cameras rely on large baselines, which can introduce significant occlusion. Shadows cast by objects and bin walls can block parts of the scene, leading to missing data and increased noise, especially when items are stacked or tightly packed.

The Zivid 3 XL250 camera capturing a depalletization scene from a distance of 4 meters at Mujin.

In depalletization applications, materials may include cardboard, reflective and transparent plastic, or metal, often stacked in a wide range of configurations, from tightly packed to loosely arranged. In these mixed-SKU scenarios, it is critical to identify a safe grasp point for a suction cup to grasp the item’s center of mass to avoid unstable or duplicate picks. Achieving this requires accurate object segmentation and a complete, high-quality point cloud.

Challenging ambient conditions further increase complexity. Depalletization zones are often located near loading bay doors, where temperature fluctuations and changing daylight conditions are common. These environmental variations can affect the camera calibration and appearance of color images, making it essential to select a 3D camera that performs reliably across varying factory conditions.

Long working distances, combined with challenging materials and variable lighting, often result in noisy point clouds and inaccuracies that hinder reliable object detection. When Mujin’s vision engineering team set out to improve robotic picking performance, three requirements were non-negotiable:

• Exceptionally accurate and stable point clouds
• Minimal point-cloud noise, even in challenging scenes
• Good quality and stable color images

These factors directly determine whether robots can consistently detect parts, avoid collisions, and maintain stable performance across thousands of picks.

03 Solution

The Zivid 3 XL250 was engineered to address these exact challenges and support large workspace automation. It enables companies like Mujin to approach demanding applications such as depalletization and large bin picking with greater confidence.

Mujin uses the Zivid 3 to enhance its warehouse and logistics automation capabilities. The camera achieves a trueness error below 0.2 percent, enabling accurate capture of 3D geometry for challenging objects, including black, shiny, or transparent materials. Its 3D HDR capability and 8 megapixel 3D data provide millimeter level accuracy at long working distances, supporting reliable object detection and precise robotic motion planning.

Zivid 3 XL250 camera

Unlike many long distance 3D systems, the Zivid 3 maintains a relatively small baseline, which helps reduce occlusion caused by shadows. It also supports all international standard pallet sizes, allowing Mujin to capture complete point clouds without shape discontinuities or large depth variations - in under 500 ms capture time. 

Stability and ease of deployment were also important considerations. Zivid’s ready-to-use presets for a wide range of applications, lighting conditions, and materials help reduce setup time. The intrinsic calibration process is simple and takes only a few minutes, while providing real-time trueness feedback to support long-term stability in changing factory environments.

04 Results

Years of close collaboration between Zivid and Mujin made the Zivid 3 possible through continuous feedback and rapid iteration. Together, the two companies moved at the pace required by demanding end customers.

With improved 3D data quality and system stability, Mujin further strengthens its ability to deliver automation solutions with fewer human interventions, as well as high throughput and reliability. The straightforward implementation of the Zivid 3 helped streamline engineering efforts and contributed to more efficient project execution and lower overall project costs.

The collaboration continues to grow: Mujin was featured as a launch customer for the Zivid 3, which was demonstrated at iREX 2025 in Tokyo. 

Mujin demonstrating high-performance multi-SKU depalletizing robots powered by the Zivid 3D camera at iRex 2026

Like Zivid, Mujin pushes beyond the standard benchmarks, aiming for success rates exceeding 99.9%. Even a single mispick can compound into substantial losses over time, making reliability a key differentiator. With a shared vision, the two companies have reached new levels of stability, robustness, and performance that were simply not achievable before.