Why Industrial Robotics need Artificial Intelligence to succeed.
As we enter the fourth industrial revolution, many factories are starting to employ industrial robots to automate their production processes, enhance some processes or gain better operational performance.
However, these robots don’t operate independently but need either humans, software, or a combination of both to bring value to the enterprise. This is where artificial intelligence (AI) comes in — augmenting humans and robots.
Industrial robotics is a subfield of robotics that deals with the design, construction, operation, and application of robots in industrial environments. These robots are used to automate tasks that are hazardous, difficult, or impossible for humans to perform.
What kind of AI is used in industrial robotics?
The most common type of Artificial Intelligence technology we see is Computer Vision, where digitally captured media such as RGB or thermal images and video are captured and analyzed against specific use cases.
The first step typically is to capture data. This can be done using static cameras and sensors, but advancements in mobile robotics such as SPOT from Boston Dynamics and drones from companies like DJI have opened up many additional use cases.
For example, using an agile quadruped robot such as SPOT allows for the automation of routine inspection tasks and data capture to be done safely, accurately, and very often more frequently than a human can.
Adding specific payloads like Spot CAM+IR adds a thermal camera to the Spot CAM+ to enable detailed thermal and visual images and video to be captured. However, if you want to automate beyond data capture, you have to use a combination of:
- Computer Vision — to analyze the images and video
- Machine Learning and Statistical Analysis — to gain insights into what the Computer Vision (CV) & Thermal Models detected
As can be seen in the example above, multiple different AI and Machine Learning models are used to create insights.
- Just capturing data is not enough.
- Just capturing data, and identifying the content using Computer Vision (CV) is not enough
You need to have either real-time and/or offline batch processing of that data using additional Machine Learning algorithms to provide insights.
For the real-time scenarios, AI really shines because if you have fully autonomous robots, you don’t want a human to still be involved in the minute-by-minute operations and inspections being conducted. You want full automation.
This is where AI can take over that filtering of information to what’s important, and alert and inform a human only when needed. This still allows a human to be involved. However, it splits the responsibility as:
- Offload data capture and filtering to the robots and AI
- Provides the human with valuable information when needed to make a decision
The analog gauge use case:
One of the first use cases we tackled was adding an analog gauge detection model to our library of Industrial Inspection AI models.
Many factories, power plants, oil and gas refineries etc., have hundreds and sometimes thousands of analog gauges. These gauges are important for operations but lack the ability to convey real-time information to a central command center for example.
Replacing all analog gauges is just not feasible, given that retrofitting digital gauges that are connected to power and the network would be cost prohibitive and the maintenance of these digital gauges are often much more involved than reliable analog gauges. In many cases, disconnected analog gauges are also more secure, for example in nuclear facilities where connected digital gauges could potentially be hacked.
This is where drones and robots can be leveraged using payloads to capture images and AI to analyze any thresholds which are used to alert humans in real-time only when needed. The data captured can also be stored and further analyzed for predictive maintenance and other Advanced Predictive Analytical use cases.
The cost? A robot and AI software, removing the need to retrofit all analog gauges.
Conclusion:
AI adds the “brains” to the robots, making them more useful than just agile remote data capture devices. AI is used to automate routine inspection tasks end to end, along with routine data capture, providing valuable insights that would otherwise be unavailable.
The cost of AI-enabled industrial robotics is a fraction of the cost of retrofitting all analog gauges for example, and many other use cases exist with similar ROI benefits attached. AI is really essential for Industrial Robotics to succeed.
