Making a robot: It’s as complicated as you thought it would be

Robots are often a wonder to watch. Whether they’re sorting garbage on a conveyor belt or playing rock music, there’s something mesmerising about the way they move — almost as if they could be alive. 

That fluidity of movement, the robot’s ability to be programmed — these are the results of decades of work and research in engineering. It took a long time to get to where we are today.

Robots have been a part of human history since the ancient days of mythology (even before we began calling them robots) but the first industrial robot designs only emerged in the 1930s. These were mostly arm-like structures, with joints that allowed them to perform motions like pulling, pushing, pressing and lifting. 

It was only in the 1950s that the first digitally operated, programmable robot — the Unimate — came into being. 

In 1961, the Unimate joined the workforce at General Motors. This particular robot is said to be “the foundation of the modern robotics industry” and there are now a variety of robot types that can be applied in a range of industrial use cases. 

What does it take to manufacture these robots today? And what processes do we follow to ensure that the right robots are used in the right situations?

Start by identifying the problem

In order to decide which type of robot is the best fit for a particular use case, one needs to ask what problem the robot is meant to solve and why a robot is being deployed in this particular scenario. 

Is the task dangerous for humans? Is it something that’s too simple and repetitive? Is it something that a robot could do with a much higher level of accuracy? These are some of the questions that need to be asked.

Robot manufacturers typically have a set of basic designs that can be customised further to meet specific requirements based on what the robots are going to be used for. 

During this design phase, there are many factors that must be taken into consideration, such as: job to be performed, speed of operation, environment in which it is operating, materials to be handled, variables in the process and whether there will be human involvement. 

On top of that, there are also potential risks that need to be anticipated and accounted for. The designer must ask questions such as: what would failure look like and what could be done to prevent it from happening?

This may sound like a lot for just the design phase but it isn’t even the most complicated part of the robot manufacturing process.

Elegant solutions require effort  

A single industrial robot may consist of 2,000 individual components like electric motors, hydraulic cylinders, bearings, wiring, controllers, batteries, axles, wheels and so on. And these components will come together to form the following parts: controller, arm, end effector, drive system and sensors. 

The controller is the “brain” of the robot. This is where the robot’s programming goes and the information here determines how the robot operates. 

The drive system is the engine that allows the robot to move. Like many other robot parts, there are many types including hydraulic, electric and pneumatic drive systems, which are selected based on what the robot will be used for. 

Sensors are like the robot’s “eyes and ears”. Combined with the right programming, they allow the robots to differentiate between objects, avoid colliding into obstacles or orientate itself. 

Sometimes sensors work closely with the end effector, which functions as a “hand”. It doesn’t always look like the five-fingered appendage that we’re so used to; sometimes it’s a gripper, a welding torch, a magnet or even a vacuum pump. 

The arm itself is an engineering marvel. Inspired by the human arm but taken a step further, robot arms also have axes, joints and links that enable the arm to twist, bend and turn in an extensive number of ways. 

Putting a machine like that together requires thoughtful design, precise assembly and finally, well-developed programming.

A robot is only as good as its algorithm

After the robot is built, it needs to be taught what to do. Some tasks may be simple and repetitive; robots doing these tasks will need to be programmed accordingly. 

This in itself is a complicated process that requires a knowledge of software development, as well as a deep understanding of robot engineering.  

However, there are other processes with a much larger number of variables that require machine learning algorithms. These algorithms allow the robot to “make decisions” based on training data so that it functions the way it is supposed to even without explicit programming. 

Take for example, a trash sorting robot that separates plastic bottles from other trash. It would be extremely tedious to explicitly program this robot. There would be too many variables to take into consideration. 

A more efficient solution would be to equip the robot with computer vision elements and a machine learning algorithm so that it can recognize and differentiate between plastic bottles and other trash.

High quality training data is needed to develop an accurate machine learning model. This is where having a partner like Supahands for data labeling is key. Although it’s possible to conduct data labeling in-house, it is often more efficient to partner with a specialist company that will be able to provide high quality, accurate training data sets.

Making sure it all works

When the robot is built and programmed, it is then installed and tested on-site. If it’s a stationary robot, it needs to be properly secured in its position. If it’s a moving robot, it needs to be tested to ensure that it moves along its designated path.

In any case, during the installation process, safety measures need to be set up to prevent accidents. For example, construction of fencing around a stationary robot so that humans don’t accidentally wander into the robot’s area of movement. 

There are also quality control measures that need to be completed. These QC tests consist of two parts: functional accuracy and burn-in

The functional accuracy serves to test specific functions of the robot within its actual position to ensure that its motions and actions are all accurate. At this stage, if there are any problems, adjustments and fixes are made. 

The robot is then made to operate for several hours during the burn-in process. This allows loss of accuracy detection, as well as check operating temperature, which may affect how accurate the sensors are. 

Making work easier for humans

Robots in the workforce are meant to make work easier for their human counterparts. This is why it’s important that the robots deployed are easy to install and use. 

Although a robot is such a complex piece of technology, its makers have always had to find innovative ways to package them for maximum user friendliness as these robots are often used by non-tech personnel. 

Most robots do not come out perfect immediately. Many times, deploying a robot involves a trial-and-error process. As the robotics industry advances further and new solutions develop, this process is something we have to get used to. It’s a requirement for innovation and will help take us to the next stage of robotics. 


Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.

You May Also Like