Powering Retail Robotics: Jabil’s Machine Vision Core

Executive Summary

Client: Jabil

Industry: Manufacturing | Robotics

Engagement: R&D Partnership

Country: United States

Project at a Glance

Jabil needed a robotics system to detect and avoid hazards in unpredictable retail environments. Existing vision technologies couldn’t identify small objects or transparent liquids, requiring a custom perception engine. VOLO developed the engine using computer vision and motion control algorithms on Linux, enabling real-time hazard detection and autonomous rerouting on limited hardware. The solution allowed the robot to adjust navigation without human oversight, reducing costs and improving safety, while setting the stage for future innovation.

400+

stores in pilot phase

floor scans per day

95%+

detection accuracy

45%

reduced manual inspection

About the Client

Jabil is a Fortune 200 manufacturing solutions provider with a global presence in advanced technologies, robotics, and intelligent automation. The company partners with leading enterprises across industries to design and deliver innovative products at scale, driving efficiency and modernization in highly competitive markets.

Jabil's vertically integrated model covers the entire product lifecycle in the supply chain management, from innovation and design to sourcing, manufacturing, delivery, and service. This comprehensive approach enables the company to provide agile, advanced manufacturing solutions with unmatched engineering and technical expertise.

The Challenge

Jabil was developing a new generation of retail robots to operate safely in unpredictable environments. The goal: enable fully autonomous navigation by detecting and avoiding hazards like spilled liquids, small objects, and debris in real time. The detected hazards were reported to the data processing center, where necessary actions were taken to mitigate possible safety risks.

The existing solution relied on built-in LIDAR, which detected only medium and large obstacles, leaving small and liquid hazards undetected. A huge R&D work was needed both in computer vision and the development of complex mathematical algorithms. The ML approach was investigated as well, though the current computational power of the robot’s hardware was not enough at all, so this approach was abolished at an early stage of development.

Key Challenges

Off-the-shelf vision systems couldn’t reliably detect floor-level hazards
Limited onboard computing power constrained advanced ML approaches
Real-time hazard detection required instant decision-making and rerouting

Beyond solving the technical challenge, Jabil’s broader objective was clear: to reduce operational costs, improve safety, and establish a foundation for autonomous systems that could transform retail operations at scale.

Have a Complex Challenge to Solve?

Talk to an Expert

Our Solution

VOLO partnered with Jabil’s R&D team to design the robot’s perception and navigation engine from the ground up.

Instead of following standard Agile cycles, the project relied on a research-first methodology, which helped the R&D team to focus on experiments and constant precision improvement. Every algorithm was developed and tested from scratch.

Although the system was designed using relatively decent hardware, the performance and power-saving aspects were still crucial for the whole ecosystem.

Custom Computer Vision Module 

VOLO developed a custom computer vision module in C++ using OpenCV to detect small objects and transparent liquids across diverse retail floor surfaces, optimized for real-time execution on constrained devices. This solution achieved consistent detection accuracy in scenarios where conventional systems struggled, directly reducing operational risks in live retail environments. Reliably identifying potential hazards, it allowed personnel to respond faster, prevented accidents, and improved overall safety on the shop floor. 

Autonomous Motion Control and Hazard Reporting 

The system included routing logic that interpreted hazard coordinates and rerouted autonomously, maintaining precision and stability even in dynamic retail spaces. By detecting hazards promptly and alerting on-site personnel, the solution enabled immediate removal of risks, protecting clients and restoring the environment to a safe state. This approach minimized safety incidents, reduced downtime, and ensured that hazard management became proactive rather than reactive. 

Server-Side Monitoring Dashboard 

All detected hazards were reported to an integrated monitoring dashboard, where on-site employees could take the necessary actions. By centralizing hazard information and automating location inspection, this system minimized response time for incidents and cut operational costs that would have been spent on manual inspections. The dashboard also provided real-time oversight, helping management track hazard trends and improve preventive measures over time.

The Impact

VOLO’s solution gave Jabil’s retail robot the ability to detect and avoid hazards in real time, transforming how autonomous systems could operate in live environments. Manual oversight was no longer required, enabling safer operations and lowering costs tied to human monitoring.

Robots could now process floor-level hazards dynamically, reroute instantly, and complete tasks with consistency, all within the constraints of low-power embedded hardware.

The result is greater autonomy, accuracy, and efficiency. By achieving this milestone, Jabil advanced its R&D in robotics and created a strong foundation for future innovations.