In the low-margin world of retail, efficiency is everything. But for Japanese robotics innovator MUSE Inc., the goal wasn't just to build a machine that could work faster than a human—it was to build one that could work with them.
Enter Armo: a collaborative autonomous mobile robot designed to solve the chronic labor shortages and physical burnout plaguing the retail industry. By handling the heavy lifting and tedious data gathering, Armo empowers human staff to focus on what they do best—customer service and creative store management.
Building a robot that can safely navigate chaotic store aisles while analyzing its environment in real-time is a massive engineering hurdle. To give Armo the "eyes" it needed without crippling its central processor, MUSE turned to Luxonis.
Meet Armo: The Multi-Purpose Retail Assistant
Unlike traditional robots built for a single task—like cleaning floors or moving boxes—Armo is designed as a versatile, modular platform. It solves three distinct problems that have long frustrated retailers:
The Heavy Lifting: Retail staff often suffer from physical strain due to restocking shelves. Armo acts as a transport assistant, autonomously hauling heavy carts from the backroom to the aisles. This reduces injury risks and keeps employees energized.
Inventory Blind Spots: Keeping track of stock is tedious and error-prone. Armo navigates the store independently, scanning shelves to detect out-of-stock items, misplaced goods, and pricing errors in real-time.
Customer Experience: Instead of wandering aisles looking for help, shoppers can interact with Armo, which guides them directly to the products they need while also serving as a mobile shelf for promotions and featured items.
To achieve this, Armo relies on the Eureka Platform, a cloud system that turns the visual data collected by the robot into actionable insights for store managers.
The Challenge: The "Vision Tax" on Performance
For Armo to be a safe and effective coworker, it needs impeccable situational awareness. It must detect humans, recognize doors, and map complex environments instantly.
However, the MUSE engineering team hit a wall during early development. Processing these high-resolution video streams and running complex AI models (like human detection and depth estimation) directly on the robot’s main CPU was a disaster for performance. The "vision tax" was too high—CPU utilization skyrocketed, causing the system to become sluggish, especially under the variable lighting conditions typical of grocery stores.
They tried other camera solutions, but the results were frustrating. Previous systems were unstable, with features that didn’t work as advertised, leading to reliability issues that simply aren't an option in a busy retail environment.
The Solution: Offloading AI to the Edge
MUSE found the answer in the Luxonis modular CCM vision ecosystem. By deploying flexible base boards paired with high-performance camera modules, they were able to completely rethink the robot's architecture.
"By offloading intensive visual processing to Luxonis’ on-board AI chip, we were able to reduce CPU utilization on the main PC," the team shared. "This allowed the system to maintain stable performance and responsiveness even when processing multiple vision tasks simultaneously."
The OAK devices now act as Armo's dedicated "visual cortex," running custom in-house computer vision models for:
- Real-time environmental perception
- Human and obstacle detection
- Door recognition and safe navigation
This shift allowed the main PC to remain cool and responsive to other tasks, ensuring Armo operates smoothly even when juggling multiple operations at once.
Seamless Integration by Design
Performance was only half the battle; the physical fit was just as important. Retail robots need to be sleek and non-intrusive, leaving little room for bulky hardware.
This is where the flexibility of the Luxonis modular design shines. It allows engineers to mix and match the perfect combination of sensors and compute capabilities—placing high-resolution sensors for detail alongside wide-angle modules for navigation—while fitting the hardware into the most convenient spots within the robot. Because the sensors connect via flexible cables rather than being locked into a rigid box, the engineering team could integrate the vision system directly into the Armo chassis. This eliminated mechanical constraints and allowed them to maintain the robot's approachable, friendly aesthetic without compromising on sensor quality.
From Prototype to Scale
The transition from a prototype to a sizable fleet of robots has been seamless. Throughout the testing and initial deployment phases, MUSE reported zero major issues with the vision system. Unlike their experience with previous vendors, the Luxonis hardware worked exactly as expected, backed by clear documentation and a stable software environment.
Why MUSE Chose Luxonis
As MUSE continues to scale their operations into the US market, their decision to standardize on Luxonis came down to four key advantages:
- On-Device Inference: The ability to offload intensive AI processing to the camera module itself was critical, freeing up the robot’s main CPU for other operational tasks.
- Modular Design: The OAK-FFC ecosystem allowed them to customize their sensor array and integrate components into a tight form factor that pre-cased cameras couldn't match.
- Proven Reliability: Unlike previous solutions that were buggy and unpredictable, the Luxonis ecosystem provided a stable hardware and software foundation that worked exactly as advertised.
Looking Ahead: A New Era for Smart Retail
The partnership between MUSE and Luxonis represents more than just a smarter robot; it signals a shift toward a truly intelligent retail environment. As MUSE accelerates its expansion into the U.S., the success of Armo proves that complex robotic vision doesn't require massive servers or bulky hardware—it just requires the right edge technology.
With Luxonis modules acting as the eyes of the operation, the possibilities for the retail industry are vast. From autonomous cleaners that can spot spills instantly, to smart shelves that recognize product engagement, to frictionless checkout systems; the future of retail is being built on flexible and low-power Edge AI. By removing the technical barriers to vision, Luxonis is helping pioneers like MUSE turn physical stores into data-rich, highly efficient spaces where technology and humans work in harmony.
Learn More
