As part of our collaboration with Tevel Aerobatics, a company developing autonomous fruit-picking robots, TandemG operated as an extended R&D Extension Team.
The primary objective was to strengthen the reliability of the embedded platform and prepare it for scalability and long-term maintainability in complex agricultural environments.
The project focused on two main tracks: system stabilization and infrastructure modernization.
On the stabilization front, we addressed a persistent software issue within a third-party image processing library that disrupted the vision pipeline and caused system instability. Through deep debugging, advanced tracing techniques, and memory allocation analysis in a Yocto-based environment, we identified a hidden logical flaw in the error-handling mechanism and implemented a targeted patch that fully resolved the issue.
In parallel, we led a comprehensive infrastructure upgrade of the operating system from Yocto Zeus to Yocto Hardknott. This effort included integrating Docker and K3S to enable a modular, containerized runtime environment, implementing NetBoot capabilities to support remote OTA updates for both root and boot partitions, and delivering custom Device Tree adaptations tailored to the client’s hardware.
Results:
- System Stability: Root-cause identification and resolution of a critical vision system issue, resulting in a stable and consistent image-processing pipeline.
- Technological Upgrade: Migration to an up-to-date Yocto platform with full container support (Docker / K3S).
- Production Readiness: Infrastructure designed for FOTA updates and improved maintainability, supporting future scale and growth.
Technologies:
- Platform: Yocto Linux (Hardknott)
- Containers: Docker, K3S
- Monitoring & Time Sync: Fluent Bit, Chrony
- Programming Languages: C / C++
- DevOps & CI/CD: GitLab, Jira Cloud
- Debugging Tools: GDB, Memory Tracing, Custom Instrumentation