Tend.ai’s Machine Learning and Cloud Services Solution Expands Robotics Role in Manufacturing

Applying machine learning (ML) and cloud services to robotics automation creates new opportunities for manufacturers to achieve higher reliability, increase productivity, and more. An example of this approach is Boston-based startup Tend.ai, which offers a novel cloud-based ML service that allows manufacturers to use articulated robotic arms to physically operate Computer Numerical Control (CNC) machines, 3D printers, and other machinery used in manufacturing as would a human worker. The company, which emerged from stealth less than a year ago, received a US$2 million seed money in November 2016.

Manufacturing and machining businesses with experience deploying robotics solutions are continually seeking ways to extend automation beyond their current configurations. One approach, the Industry 4.0 model for digitalizing manufacturing businesses, produced tangible benefits such as lower production costs, as well as higher quality output. However, Industry 4.0 projects require an enormous amount of time and experience to connect machines to communicate with other, limiting Industry 4.0 opportunities for small, less advanced firms. In addition, the complexities of workflow and programming often limited production scale initiatives to simpler pilot programs.

Tend.ai provides a simplified, low-cost, digitalized solution that is in keeping with the tenants of Industry 4.0 for the common manual manufacturing task of tending machines such as 3D printers and CNC machines. It combines ML, the cloud, and robotics (including those collaborative bots capable of working side-by-side with humans) to deliver an entirely software-based solution.

With the Tend.ai solution, web cameras provide the vision for the robotic arms by reading the LED status screens on CNC machines and 3D printers. The cameras can interpret where the machine is in the manufacturing process by sending the image to the cloud for processing. When an interaction is desired, the cloud sends back commands for the robotic arm to act accordingly—pushing buttons on the machine, opening the machine cover, and removing completed parts. The system is also capable of handling error messages. By incorporating vison into its solution, the Tend.ai system eliminates the prohibitive requirement of a machine-to-machine network.

Figure 1: A Universal Robots' UR3 Operates Multiple 3D Printers

Source: Techcrunch

Computer vision and ML are critical to what makes Tend.ai’s cloud-based solution work. The ability to read machine displays, as well as locate and physically push buttons, is vital to the operation of machines in a human-like manner. Alternative techniques such as optical character recognition (OCR) and other scanning tools proved to be unreliable.

The training data generated by Tend.ai’s customers for specific machines is captured and stored using cloud services. These specifications can then be reused and made available through the cloud. In many cases, individual robotics users no longer must train their own systems. The Tend.ai solution works out of the box. In addition, as machine vendors roll out new software and as new error codes surface for their hardware, Tend.ai customers get the benefit of learning from each new deployment. As greater awareness of the cloud model for “train once, deploy to all” takes hold, machine manufacturers are likely to become involved in the training process so their customers have little to no downtime associated with user interface (UI) updates and new machine deployments.

For more in-depth information on robotics markets and intelligent systems trends, see the ABI Research service Robotics, Automation & Intelligent Systems.