IFOY Audit
Finalists undergo the three-stage IFOY Audit as part of the IFOY Audit, which takes place during the TEST CAMP INTRALOGISTICS. This consists of the scientific IFOY Innovation Check, the IFOY Test and the jury test.
Semantic visual intelligence for intralogistics
Romb Technologies
By transforming raw visual data into meaningful insights, semantic perception enables machines to truly understand and interpret their environment – enabling safer, smarter, and more efficient intralogistics operations.
Product name and company
Semantic visual intelligence for intralogistics – Romb Technologies
Description
- Processes video streams or static images in real time.
- A semantic segmentation model (based on machine learning) identifies relevant object classes – such as pallets, loads, transport corridors, vehicles, and people – and their precise positions in each frame.
- The segmented images can be enriched with facility infrastructure data to enable real-time monitoring, e.g., detecting loaded or empty pallet positions, misplaced items, or equipment
- Depth integration: by fusing data from depth cameras or LiDAR, the system can accurately recognize the position and shape of pallets and detect obstacles in the transport areas enabling advanced functionalities, such as obstacle-specific avoidance behavior (e.g., avoid static obstacles, stop near persons) or adaptive pickup of imprecisely placed pallets.
- The model is trained on intralogistics-specific datasets, ensuring robust performance in warehouse environments.
- Can be fine-tuned for specific facilities, further improving site-specific precision and adaptability.
Innovations
The technology gives machines human-like environmental understanding. Unlike traditional vision systems limited to fixed object templates, it recognizes and distinguishes between people, pallets, loads, vehicles, and navigable terrain – unlocking new capabilities such as adaptive load handling, obstacle avoidance, and real-time facility-level analytics.
Key innovations include:
- Generic object perception, independent of pallet or load type.
- Automated camera-fork calibration for load handling (EP24223684.2 patent pending).
- Real-time operation, up to 20 fps.
Together, these features enable flexible, vision-driven automation that dramatically enhances AGV/AMR performance and facility visibility.
Market relevance
The innovation primarily targets AGV and AMR manufacturers as well as system integrators. By integrating our semantic perception modules, they can offer safer, more autonomous, and more efficient vehicles to end users in manufacturing and logistics.
The combined AGV/AMR market is estimated at about 70,000 units sold per year in 2024, and expected to grow to 110,000 units by 2030. With the expected target price for a semantic perception software package ranging between €1,000 and €5,000, the estimated market potential today is between €70 million and €350 million, and expected to grow by about 50 % by 2030.
A secondary market includes end users who can deploy the software on existing camera infrastructure to gain real-time insights into asset utilization and material flow. This segment could exceed the primary market in value as facilities increasingly pursue digital twins and AI-driven analytics.
Customer benefits
The system increases operational transparency, efficiency, and safety within intralogistics environments through the following specific benefits:
- Real-time visibility of facility assets and material flows.
- Reduced downtime by early detection of obstacles and irregularities in transport corridors.
- Simplified handovers between workers and AGVs/AMRs through contextual scene understanding.
- Automated handling of empty pallets, including collection, stacking, and destacking.
- Continuous monitoring of irregular conditions such as misplaced items or equipment.
- Reduced deployment time and integration complexity of AGV/AMR systems due to increased environmental awareness of the vehicles.
IFOY Innovation Check
Functionality / Type of implementation
The Semantic Visual Intelligence for Intralogistics from Romb Technologies is a modular solution for mobile robot systems in intralogistics that can semantically recognize people and a wide variety of objects in the warehouse (categories such as drivable surfaces, various types of full/empty pallets, pallet fronts, type of load, walls, columns, et cetera) and can trigger a correspondingly adapted robot behavior. The solution is based on state-of-the-art semantic segmentation techniques and AI model architectures and processes both RGB color image data and 3D sensor data from any sensor manufacturer, in each case based on a single frame (no tracking). Particular emphasis was placed on the low requirement for real annotated training data for the categories to be recognized: only about 20 sample images are needed per category. A dataset of this kind, comprising approximately 500 images, can be adaptively retrained by the system for a new customer within five to eight hours. For smooth operation, the system places only moderate demands on, for example, simple, embedded GPU computing hardware. In testing, the system reliably recognized the various categories of warehouse entities and people in most cases, with only a few explainable error situations and minimal fluctuation in the semantically segmented image regions.
Novelty / Innovation
The system’s unique feature lies in its simple and flexible adaptability to new customer use cases, requiring only a small amount of annotated real-world data for fine-tuning. As a modular system, it is highly versatile. Additionally, the system can automatically calibrate its installation position on vehicles equipped with forks. From a technical standpoint, the degree of novelty is moderate, as comparable systems are already in use in some cases in warehouse monitoring systems, cargo measurement systems, and autonomous forklifts.
Customer benefit
The primary benefits for customers include more efficient use of AMRs through situation-adaptive obstacle avoidance behavior, automatic pallet handling, more efficient commissioning, and the implementation of a wide range of new autonomous applications with AMRs. Both existing vehicles and new AMRs can be easily equipped with Romb’s plug-and-play system. Thanks to its modularity and low target price, the system is a worthwhile investment in many cases. In addition, the short commissioning time of just one to two days significantly simplifies deployment.
Market relevance
The AMR market is a significant growth market, with approximately 70,000 units expected to be sold by 2024 and is increasingly opening up to more complex and demanding application scenarios. A semantic understanding of the environment is a particularly valuable feature for part of the fleet in such situations. To tap into the market, Romb is pursuing a licensing model with AMR manufacturers and is very competitively priced at 1000 to 5000 euros per system. To date, three different application scenarios have been piloted with one end customer and two interested manufacturers: the adaptive, autonomous picking of pallets, adaptive avoidance behavior depending on obstacle types (people, static objects), and stock availability checks in a buffer rack.
IFOY verdict
Romb presents a technically simple yet high-performance and compelling solution for semantic environmental understanding for AMRs, which stands out for its highly data-efficient training. Scaling and market penetration represent the next critical steps for the company.
| Functionality / type of implementation | + |
| Novelty / Innovation | Ø |
| Customer benefits | + |
| Market relevance | + |
| ++ very good / + good / Ø balanced / – less / — not available | |
Videos
Playlist
1 Videos
Playlist
4 Videos
Image/video credits: Romb Technologies
Discover more
Statutes
The bylaws set forth the core values and standards of the IFOY AWARD, which guide the organization’s actions. These include, among other things, governing bodies, the election process, evaluation criteria, and the Code of Conduct.
Committees
The International Intralogistics and Forklift Truck of the Year has three bodies: Jury, sponsors, and advisory board.