Modern Warehouse Monitoring systems integrate a variety of advanced technologies to achieve efficient, automated warehouse operations. These systems include Internet of Things (IoT) sensors, big data analytics and artificial intelligence (AI), warehouse execution systems (WES), automated storage and retrieval systems (AS/RS), cloud computing and data storage, as well as security and monitoring systems. Through the collaborative work of these technologies, warehouses can achieve real-time data collection and analysis, optimize resource allocation, reduce manual intervention, and improve operational efficiency.
Among these systems, the accuracy of warehouse monitoring is particularly important. By adopting LiDAR or vision technology, it is possible to monitor each inventory slot’s occupancy status and stacking information in real-time. This lets mobile robots obtain real-time, detailed data support for the next accurate wares handling operation. However, the use of warehouse monitoring still faces the following challenges:
Challenge 1: AGVs and manual operations coexisting in the same warehouse cause WMS system updates to be untimely
Currently, in most warehouses where AGVs have been implemented, there is still manual stacking of items. This situation can cause the WMS system to be unable to accurately determine the actual occupancy status of each inventory slot in real time, leading to inaccurate information being provided to the AGV. Suppose the occupied inventory slots are not identified promptly. In that case, it can result in erroneous instructions being received by the AGV, reducing operational efficiency and increasing the risk of collisions with other goods or equipment. In severe cases, this could lead to safety accidents, endangering personnel safety and equipment operation.
Challenge 2: Single-point LiDAR-transmitted warehouse monitoring information lacks accuracy, causing stacking accidents
Some warehouses use single-point LiDAR for warehouse monitoring. This type of rangefinder can only emit a single laser beam to the surface of an object to form a point. This measurement method may sometimes overlook the gaps between boxes or pallets, leading to the misidentification of the inventory slot as empty or lacking goods, thus easily causing stacking accidents.
Challenge 3: Sole reliance on RGB cameras for validation methods poses risks of misjudgment and information loss
Using RGB cameras for determining the status of inventory slots has several drawbacks. Firstly, when detecting target areas through deep learning, the presence of objects outside the training set entering the inventory slot can easily cause misdetections, providing incorrect inventory slot status information. Secondly, for inventory slots that require cargo stacking, the information provided by the RGB camera lacks cargo height information, making it difficult to arrange stacking tasks. Additionally, some warehouses use ultra-wide-angle fisheye cameras to determine inventory slot status, which suffers from significant edge distortion, presenting challenges to model training and prediction accuracy, and requiring additional GPU server costs.
MRDVS Cost-Effective 3D Vision Solutions Applied in Multiple Warehouse Monitoring Projects

MRDVS, a subsidiary of Lanxin Robotics, is dedicated to providing advanced vision solutions for mobile robots and their associated warehouse and logistics scenarios. Through continuous innovation, MRDVS enhances automation levels and operational efficiency in the industry. Since the introduction of its 3D vision-based inventory slot detection solution, MRDVS has been used in multiple warehouse projects in conjunction with AGVs, including in the lithium battery, logistics warehousing, and packaging industries.
This solution is developed based on RGB-D cameras, which collect three-dimensional data and color information of objects stacked on each inventory slot. Combining AI deep learning technology, the recognition algorithm is placed at the camera end to achieve comprehensive monitoring of warehouse status, minimizing the possibility of misjudgment. Moreover, MRDVS provides a complete set of solutions from semi-automatic annotation to model training and deployment. Users can flexibly enhance the model while greatly reducing the difficulty of data annotation.
By focusing on warehouse monitoring and slot detection for warehouse automation, MRDVS ensures accurate inventory space detection and enhances the overall efficiency and safety of warehouse operations.