Why Robots Should Be Open-Architecture and Hardware-Agnostic

For industrial adoption of flexible AI solutions to make good business sense, industrial users need to know they can maintain the system on-site without having to rely on the access to rare specialists. The lack of transparency in black box robotic systems hides the underlying algorithms and decision-making processes, making it difficult for users to troubleshoot malfunctions or optimize performance. Open-architecture, hardware-agnostic solutions allow for greater visibility and flexibility in how robots are programmed and operated while offering hiring and sourcing flexibility that reduces cost and risk.

Vendor lock-in is a significant risk with proprietary solutions that require specific hardware to match the software. Businesses may find themselves tied to a single vendor for support, software updates and hardware replacements, restraining their capacity to prevent downtime if supply chain issues exist. Proprietary solutions also restrict the sources of expert knowledge, customization and the ability to innovate or integrate newer, potentially better hardware as requirements change. Open-architecture robotics solutions reduce sourcing risks and offer flexibility, allowing companies to adopt the best tools for their needs without being confined to one vendor's ecosystem or reliant on a single supply chain.

Robotic pick-and-pack systems: a key industrial application

Picking has always been one of the most costly and labor-intensive activities in warehousing and production processes. Nearly all online orders in e-commerce or manufacturing are picked and packed by human hands today, representing a significant portion of operating costs. At the same time, the industry’s labor shortage continues as consumer demands and competition increase.

Robots are already indispensable in many areas of industry, but for some tasks they are far too inflexible. Pick-and-pack systems, for example, require adaptive, accurate operations when dealing with unknown or unpredictable objects and cannot rely on traditional robotics deployments designed to repeat the same task. These processes require the complex logic and AI models that can handle many variations within a specific task with the necessary accuracy and quality control. This process complexity is coupled with an increasing demand to expand operations, as pick-and-pack throughput volume is expected to increase 40% globally by 2030.

The proprietary SaaS problem

Robotic solutions that allow both increased throughput and cost-efficient scaling are the obvious answer. So why aren’t more AI-driven robotic solutions being employed? The adoption of these technologies has been hamstrung by the proliferation of proprietary software-as-a-service (SAAS) solutions that require solution-specific expertise and hardware, driving up customization costs and creating single-source and downtime risks.

Projects must have tight margins for cost and risk to successfully deploy pick-and-pack solutions in complex environments. Unexpected customization costs, for example, can drive a project far beyond its scope and invalidate the projections for expected ROI. Hardware choices must fit within project cost limitations and have reliable supply chains for maintenance and future operational scaling on demand. Software solutions should allow easy customization by any engineer, not specialized high-cost consultants with limited availability. On-site deployments are necessitated to limit exposure to cybersecurity risks and future-proof operations against downtime due to server issues caused by ISP outages, natural disasters or even the acquisition of, or cessation of support by, the original vendor.

Open-architecture, on-site robotic AI solution architectures based on standardized technologies allow any hardware to be used based on application needs and project cost limitations, ensuring that robotics solutions can be deployed as needed and scaled to meet demand. Since implementations must be able to employ the vision and robotic hardware components that the project requires, modular open-architecture solutions with easily adaptable applications and extensions offer flexibility compared to solutions confined to the hardware offerings and supply chain of the AI provider.

Using standardized technologies means that open-architecture solutions should allow for application-specific customizations while using existing IT and OT standards and market-proven, commercially tested technology for visualization, customization and data-sharing. For example, out-of-the-box pre-trained perception AI for robots must enable users to push the boundaries of their current use cases by leveraging modern software development methods for automation development while maintaining vendor-independent robot programming using industry standards, such as the Standard Robot Command Interface (SRCI). Working within a standardized automation platform enables most engineers to create AI-driven robot solutions instead of relying on a limited number of specialists available for niche proprietary systems.

Enabling varied supply chains while minimizing risk

The adoption of AI-driven robotic pick-and-pack systems will be essential over the next decade to increase efficiency, reduce costs and meet rising demand. This will be the case across numerous industries. Modular, open-architecture AI systems that enable varied supply chains and minimize risk are required to facilitate this evolution. Proprietary, closed and cloud-based systems will only slow the adoption of highly efficient robotic solutions powered by AI and compartmentalize robotics and AI expertise, further exacerbating the shortage of critical specialized knowledge in the labor market.

Allowing industries to maintain control over their operations, use affordable and widely available hardware and expertise, and quickly scale as needed will encourage adoption and increase competitive advantage alongside efficiency. Once industrial adoption is widespread, the cost savings across the board — coupled with improved process efficiency — can drive lower prices, improving markets and the economy, while creating better business environments for everyone.

Eugen Solowjow is R&D manager of AI and robotics at Siemens.