Produce Alignment Automation
Motivo Partnered with a Leading Food Processor to Automate and Streamline Onion Alignment
Background
Traditional onion alignment is a labor-intensive and inconsistent process. A leading food processor, a major player in the wholesale/distribution of fresh fruits and vegetables in the USA, sought to gain a competitive advantage while improving efficiency and reducing costs.
The work environment itself was harsh and unpleasant for human workers; onions have to be refrigerated, so workers had to wear uncomfortable, layered protective gear to keep themselves warm and to combat the pungent fumes that caused irritation to the eyes and nose. This made it difficult to staff the production line and contributed to a desire for reduced labor costs and improved employee retention. Existing solutions in the market did not satisfy the client’s requirements.
The project demanded a robust solution capable of handling inconsistent produce, as onions vary significantly in appearance, shape, and size. The appearance of the onion, including its variation in shape and colors, also presented specific challenges for the development of machine vision. It needed to achieve smart and precise alignment consistently and efficiently, while also withstanding the harsh conditions of a processing plant, including corrosion, water intrusion, and challenging cleaning processes. Furthermore, the solution had to integrate seamlessly with existing high-speed processing lines without requiring a complete overhaul.
Concept to Functional Unit
Motivo collaborated closely with the client throughout the development process, which began with an initial site visit where the problem was broken down into different aspects and tackled sequentially, building from stage to stage into prototypes that were more and more functional and higher fidelity.
Motivo started by brainstorming and rapidly testing many different concepts for how to manipulate the onions and accommodate the slippery onion paper and all the various shapes and sizes without damaging the produce.
CAD Image of Early Bench-Test Assembly
CAD Image of First Machine Used on Production Line
A proof-of-concept model was built utilizing Motivo's in-house laser cutting and 3D printing capabilities. This prototype incorporated the actual alignment motion using stepper motors on two sets of clamping arms, enabling motion on two axes.
Proof of Concept Model
After proving the handoff between the sets of grippers, the team moved to one of the most challenging aspects: machine vision. This entailed adding a camera and lighting to the prototype and undertaking extensive development work to create a highly effective custom neural network that would identify the roots and tips of onions. The positions identified by the neural network were then fed into an algorithm that informed the machine how much rotation was required, and on which axes, for accurate and consistent onion alignment.
This was a groundbreaking application of artificial intelligence at the time, predating the "AI boom," and positioning Motivo as an early innovator in deploying advanced AI for industrial robotics. Motivo built a full sensing architecture around leading object detection algorithms, as external vendor solutions proved insufficient.
Neural Network Tagging / Training Image
Neural Network Operational / Evaluation Image
Along the way, functionality was added to allow the prototype to raise and lower while the conveyor belt indexed. Once excellent performance on alignment was achieved, the project moved to the speed round, where a representative prototype was made to move at the full speed needed for production. Once that was proved out, Motivo scaled the solution, first to four lanes for an alpha prototype which was used in the production facility to gather data.
The learnings from the alpha prototype informed the final design, which aligned onions on all eight lanes per machine, achieving the full realization of the original vision for fully-mechanized onion processing.
1st Alpha Prototype
Following successful implementation of the prototype, two additional functional units were created.
Final Iteration
Outcome
Soft-touch, ultra-high-speed manipulation capabilities were developed for delicate handling and increased throughput, enabling the system to process up to 216 onions per minute per machine.
The automated alignment solution consistently performed above 90% accuracy, exceeding human performance, while operating on a production line in the processing facility.
Capabilities Used
Industrial Design: User research; Concept development and analog/digital visualization.
Systems Engineering: Complete system architecture design; Development roadmap creation; Vendor coordination and evaluation for vision systems.
Software Design & Engineering: Neural network machine vision integration; Full sensing architecture development; Controls and software engineering; Hardware and software integration.
Mechanical Design & Engineering: Electromechanical design; Soft-touch, ultra-high-speed manipulation systems; Robust design for corrosive environments; Hydraulic and pneumatic systems design.
Prototyping, Fabrication & Assembly: Rapid prototyping; Custom test bench build; Prototype implementation and testing on a production line; Manufacturing, on-site installation & post-delivery support; Low-volume production of additional functional units; Design for Manufacturing (DFM).
Testing & Validation: Performance benchmarking; Comprehensive verification and validation.