September 03, 2026
https://embeddedcomputing.com/application/edge-ai/aetina-and-moss-robotics-advance-farm-inventory-accuracy-with-edge-ai-agtech
Sponsored Blog
It’s a challenging time for the agricultural sector. Amid rising input costs, labor shortages, climate disruption, and macroeconomic uncertainty, efficiency is more important than ever. To remain profitable, specialist growers must boost yield and cut waste.
One clear way to achieve this is by replacing inefficient processes that are still widespread in the sector. For example, most tree nurseries and specialty crop growers count and grade inventory by hand, which is labor-intensive, time-consuming, and imprecise. But while automation might seem like the obvious solution, industry adoption of smart farming technology is uneven—driven in part by concerns over investment costs and ROI.
The agtech company moss robotics set out to solve the problem of manual processes—and allay the concerns of hesitant adopters—by building a cost-effective, edge AI-based agricultural robotics system for farm inventory management.
But doing so would be challenging. To fully replace manual inventory, a precision-agriculture inventory system must do far more than just count plants. The complexity of any such system would require extensive computing capabilities. moss robotics also needed an experienced engineering partner to streamline solution development.
To meet its needs, the company chose the Aetina AIB-MX23, powered by NVIDIA Jetson AGX Orin 64GB.
The Challenge: Finding compute to run multiple AI models at the rugged edge
The problem that moss robotics needed to solve was clear. Manual inventory means workers literally have to walk a field row by row, counting and grading plants by hand and recording results with pen and paper. On a large farm, it can take multiple crews months to inventory the plants for a single growing season.
In addition, manual inventory costs more than just time and labor, because it is untraceable and unreliable. Growers lack key per-plant data such as the location and maturity, and inventory is often inaccurate due to human error. Bad data can result in ~20% of production value being lost. And operationally, having no clear picture of what is actually in the field has even greater financial impact. Decisions related to forecasting, harvest, sales, and labor can only be made on a “best guess” basis—resulting in losses up to ~30% of yield.
moss wanted to offer growers something far more efficient and effective: an edge AI-powered multi-sensor kit that could be mounted on almost any farm vehicle to automate inventory management. Growers would be able to essentially drive the kit up and down rows of plants, gathering precise data on count, height, caliper (diameter), and size grading. In addition, individual plants would be geo-located with centimeter precision and recorded to a digital twin of the farm.
The company’s vision required a computing platform that met multiple demands:
- Powerful enough to run multiple AI models for perception, detection, and measurement completely at the edge to enable deployments in remote agricultural areas.
- The capacity to handle sensor fusion tasks for multiple data streams.
- Compact enough for the final sensor kit to be deployed on a farm vehicle without the need for additional equipment.
- Rugged design to operate reliably in physically demanding agricultural settings.
Finally, moss wasn’t looking only for a hardware vendor. It needed a true engineering partner to help integrate the robot “brain” with the rest of its solution.
The Solution: Aetina delivers powerful edge AI performance and multimodal sensor fusion
moss robotics selected the Aetina AIB-MX23 edge AI inference system as the basis for its solution. Based on the NVIDIA® Jetson AGX Orin™ system-on-module, the AIB-MX23 is powered by a 2048-core Ampere GPU. The platform delivers up to 275 TOPS of AI performance with 64GB of 256-bit LPDDR5 at 204.8GB/s—ample computing power and memory to run heavy edge AI workloads concurrently.
The AIB-MX23 provides simultaneous processing for multiple sensor data streams at the edge, enabling the moss solution to deliver a genuine replacement for inventories done by hand:
- Multiple cameras enable visual detection of plants for counting and size grading.
- 3D LiDAR sensors assess volumetric geometry, including height and caliper measurements.
- Real-time kinematic (RTK) GPS offers precision geolocation of individual plants.
- An inertial measurement unit (IMU) ensures accurate drive-by scanning through motion correction.
The system’s compact form factor (131 x 120 x 62.9mm or 5.16 x 4.72 x 2.47in) makes it easy to integrate into space-sensitive applications like the moss robotics sensor kit. And with an operating temperature of -25°C to +80°C, anti-shock features, and vibration resistance, the unit is rugged enough for nearly any agricultural setting.
moss robotics also found the engineering partner it was looking for in Aetina. Aetina’s edge platforms power deployments in robotics, industrial automation, and agriculture around the world, with more than 600 AI application cases in recent years. Plus, Aetina is an NVIDIA Elite Partner, with deep NVIDIA ecosystem expertise.
Its extensive edge AI and NVIDIA development experience allowed Aetina to support moss throughout the solution design process. In particular, it enabled the company to benefit from NVIDIA technologies supported by AIB-MX23 in order to:
- Streamline development through embedded and AI acceleration toolkits and libraries with the NVIDIA JetPack™.
- Run multiple edge AI workloads at the same time with NVIDIA® CUDA® and NVIDIA® TensorRT™, which offer parallel compute capabilities and inference acceleration.
| “We’ve loved working with the team at Aetina, who have been attentive and deeply involved in our success here at moss by answering emails at odd hours, taking phone calls, and advocating for us internally. The team is the main reason why working with Aetina has been such a great experience so far.”— Pravir Das, Operations Lead, moss robotics |
The Results: Proven performance and measurable ROI in real-world deployment
The result, as demonstrated by moss robotics’ real-world deployments, is an inventory process that is exceptionally faster and more reliable than manual inventory.
Installed on standard farm vehicles, the moss robotics sensor kit travels up and down plant rows at up to 10mph/16kph, performing high-speed autonomous scanning of a farm’s inventory and syncing the data to a web-based platform via cellular.
Inventory data is taken in a matter of hours instead of months. The system is highly precise as well. In a real-world deployment of ~20 units, it achieves a validated accuracy rate of 98.84% across more than 327,000 scanned trees.
That level of efficiency, accuracy, and granular data delivers undeniable ROI to growers. Inventory labor costs in the moss deployment fell by 80–95%. The system’s increased visibility into the size and grading of individual plants also enables better decisions around harvest, pricing, and sales timing—leading to an estimated revenue increase of up to 20%.
From factory to farm: Delivering on the promise of agtech
To solve agricultural sector challenges through innovation, OEMs must make smart farming technology practical and profitable for growers. Partnerships between OEMs like moss robotics and hardware experts like Aetina and NVIDIA show how it’s possible to build systems that solve real-world problems while delivering a clear return on investment.
Looking ahead, Aetina’s close relationship with NVIDIA gives moss robotics a natural pathway to product development and upgrades. Aetina’s next-generation platforms, like its AIB-AT78/68, built on NVIDIA® Jetson Thor™, will enable the company to enhance its farm inventory solution. Richer sensor fusion and support for more demanding edge AI workloads will enable higher-resolution scanning and more advanced real-time analytics.
The continuing partnership between Aetina and moss pushes the boundaries of smart farming forward, whatever challenges the future may hold.
About moss robotics
moss builds AI-powered sensor kits that automate inventory for tree nurseries and specialty-crop growers. To learn more about how moss robotics is supporting growers with data-driven truth, visit the company’s website.