Succeeding with your first deployment of autonomous inspection robots

One of the main objectives of the first deployment with an inspection robot is getting your organization ready for robotics. A properly scoped deployment will contribute to engaging the field workers and provide insights on what it takes to keep robots operational on a day-to-day basis. This will lay the foundation for high return on investment for all the following deployments. If the scoping fails, there is a risk of losing trust in robots and gaining less willingness to change in your organization. But fear not. In this post, we’re summarizing our best practices when assisting our customers to scope out their first deployment of an inspection robot. 

1. Selecting the appropriate area of deployment

It is common that there is a need to deploy robots in harmful or remote environments. These can be areas that may be highly corrosive, have a risk of explosive gases, have powerful magnetic fields, radiation from heat or nuclear etc. When selecting the appropriate robot for a plant, we recommend keeping it simple by starting easy. Select a smaller area of a plant that is to be monitored. Try to avoid external factors that might affect robot performance and sensors, such as ex-classified areas, areas with lots of humidity, sand or dust or platforms that are subject to much movement due to waves. With this approach it is easy to introduce harder challenges and harsher environments over time, which improves the likelihood of successful robot endeavours. 

In addition, it is good practise that the area has good network coverage. The minimal requirement should be to have good network connectivity at the location of the robots docking station. 

2. Selecting the appropriate robot platform

Selecting the correct robot platform is crucial. Some robots are ready for day-to-day operations while others are still immature.
The minimum requirements should be:

• That the robot can be provided with a docking station and auto-docking capabilities

• That the data captured by the robot can be extracted from the robot programmatically

3. Selecting the right use case

When you have the area and the robot, you need to select a suitable use case. Routine inspection missions are great as they can be set up once, and scheduled to run as many times as necessary at any time. To experience how fast you can get going with routine missions will contribute to providing confidence in the organization that robots can be used in day to day operations with minimal assistance required.

Common use cases for routine inspections are:

• Capture 360 images of the plant regularly for remote visual inspection

• Train simple inspection tasks that should be performed on a daily basis, such as inspecting analog gauge values, valve status etc.

4. Creating a data management plan

Selecting the appropriate data management plan is also important. It is what you do with the robotic data that enable value capture. And as much of this pipeline should be without the need of human interaction. Using various modules in Cognite Data Fusion it is possible to automate several of these pipelines from day one of operations. 

Relevant pipelines that are ready out of the box:

• Automatically map 360 images in a 3D model / 2D map to create a street view experience when remotely inspecting the plant

• Evaluating images based on existing computer vision models such as people detection and blurring, liquid spills, gauges (dial, level, digital), lever valves and more

  • Converting the imagery data to time series data

  • Evaluating the time series data against other operational data

  • Adding monitoring tasks to receive alerts when there is an anomaly or a threshold breach

• Labelling images from a specific location to begin collecting data for future computer vision models

5. Engaging the right people

By having the right people involved in the process and giving ownership to the robot and use cases will minimize the time it takes until the robot is in operation. 

In one of our most successful robot deployments, the Head of Robotics chose not to operate the robot himself. His reasoning was that “The robot is not for me - it is for the field workers. Hence, they should be the ones using it.”

In the robot deployment team, roles should be clearly defined beforehand and what people are involved should be thoroughly selected. In every robot deployment, we advice to have the following combination of people involved: 

• Domain expert: The person who normally performs inspection rounds and has the complete overview of what to look for at a specific plant. This can also be the same person that should be responsible for training the various inspection missions.

• Maintenance worker: Depending on the environment that the robot is deployed in, there may be a need for regular checks and maintenance. 

• HSE responsible: This person should be responsible for both general robot operation in the field and ensure that operators use it properly during training and mobilization.

• Operation Support central: It is likely that the operator central will be responsible for executing various inspection tasks by the robot, while also monitoring the robot while it's in operation.

• Robot-specific expert: If this is an organization's first meeting with a specific robot, someone who has long experience with it should be present. That way you will avoid spending hours or days finding a solution to 5 minute problems.

• Plant networking expert: Being able to remotely operate the robot and extracting the inspection data from the robot will require that the various ways of communicating are properly configured. Be prepared to make modifications before and during the deployment.

6. Expectation management

If you work with robotics daily, your perception of what robots are capable of is very different from someone who has no experience with robots. It is therefore important to continuously work on aligning expectations. We have listed some relevant topics below:

• What is the value gain?

  • Are you doing this deployment to keep people out of harm's way, lowering greenhouse gas emissions or to save time?

  • Is this a test project to mature the organization for all future robotic deployments?

• What are the long term goals

  • Normally unmanned operation 

  • Centralize the human workforce

  • What is the robot doing today and what is it expected to do within a year?

• What is the robots capabilities and limitations

  • Should the robot be able to climb stairs?

  • Can it avoid colliding into people?

  • What is the advantages, and what are the disadvantages

If you have other experiences or something to add to this list, please leave a comment and share with us. This also accounts for potential questions you may have.