AI-driven robots in industry: potentials and applications

Automation
AI-driven robots in industry: potentials and applications

2024-08-16

From

Maximilian Mutschler* | Translated by AI


5 min Reading Time

Where traditional image processing reaches its limits, artificial intelligence can often help. Together with robotics, this can automate processes that until now could only be performed by humans.

Artificial Intelligence (AI) has established itself as a permanent part of the industrial discussion. According to a Bitkom study from May 2024, 78 percent of German industrial companies see AI as playing a decisive role in competition. However, currently only 15 percent of companies use AI, while 53 percent wait to see what experiences other companies have with the use of AI. Almost half of the companies also bemoan the lack of know-how for the integration of AI.

60 percent of Germans see according to a survey by Eco, great potential in industrial manufacturing to overcome the shortage of labor. Robotics in particular is recognized as a promising area of application for AI. So how can industrial companies dare to integrate AI into the automation of their production processes?

AI-driven robotics: What is it really?

AI-controlled robots are industrial or collaborative robots equipped with AI vision software and cameras. This combination enables the perception of the environment and autonomous problem solving. Instead of strictly following programmed sequences, robots with AI software can simply be shown what to do. The AI then learns not to mimic the task down to the millimeter, but abstracts and performs movements that correspond to the purpose of the work step and adapt to deviations, for example in the shape or position of a workpiece. The AI can also handle color, background and light variances.

Where are the potential applications of AI in robotics?

Particularly in areas where classic automation reaches its limits, the use of AI robots is worthwhile. Typical fields of application are manual workplaces with unpredictable changes that classic robot programming cannot handle and where even special machines and systems reach their limits or are simply too expensive.

AI-driven robots also take on tasks where traditional image processing systems fail due to unstable lighting conditions or irregular parts. Examples include inserting flexible cables, assembling memory modules, screwing components together, or equipping racks.

Practical example: 
Automotive supplier screws seats with AI vision

The usefulness of AI-controlled robots in practice is demonstrated in the assembly of car seats at a German automotive supplier’s factory. The process is as follows: The screws are manually inserted into the threads by workers. Then the seats arrive in front of two robots intended to tighten the screws. However, the position of the screws varies because they were inserted manually. In addition, the seats move on a conveyor belt designed for humans, which leads to further unpredictable changes in the process. Unstable lighting conditions pose an additional challenge, rendering traditional image processing systems useless. At the supplier, Cobots, controlled by the AI software ‘Mirai’, now solve this problem. They handle variations in the position of the screws and lighting conditions with a success rate of almost 100 percent.

Identification of a use case

The example illustrates: By understanding individual sub-processes in detail, the ideal point of use for AI can be identified. In this case, it is the last step, the application of the torque wrench and secure screwing, which could not be reliably automated so far. The AI software was initially implemented in a single, previously manually performed process step. Subsequently, additional steps can be evaluated and possibly automated, such as the insertion of screws into the threads.

Manufacturers do not necessarily have to look for specific, already defined use cases such as pick-and-place or cable insertion. Rather, they should look for processes where individual, human hand movements are necessary and those that can be demonstrated to a robot.

What are the prerequisites for using AI?

Modern AI vision technologies require less specialized expertise compared to traditional robots or image processing systems. However, the success of their use depends on various factors: the commitment of management and internal technicians and engineers, the openness of employees to the new technology, and the available capacity to become familiar with the technology.

Initial experience in the field of automation and thorough preparation are also crucial. It is advantageous if robots are already in operation in the company. The automotive supplier had already experienced with automation, which enabled it to automate this application in a few days with the aid of AI. It is also important to set realistic demands on AI technologies and not to view them as a magical solution to all problems.

What does the step-by-step implementation look like?

Analysis of the application and sources of variance: The first step is to analyze the application and the sources of variance that can occur. In the example, the application is tightening the screws, the variances are the different positions of the screws and the changing light conditions.

Training the robot using the camera: After the robot has been equipped with the necessary hardware, including the camera, it is familiarized with the movement and possible variances. The camera records the scenes in which different positions of the screws occur, as well as scenarios with and without light.

Processing of data by an AI algorithm: An AI algorithm processes the data and calculates a so-called “skill” that tells the robot how to behave during execution.

Commissioning of the AI-controlled robot: The AI-controlled robot can be put into operation and reliably perform the desired action, regardless of any variances that occur. It can even handle completely new scenarios that the camera on the robot has never seen before.

Dare to take the step towards the future

Robots equipped with AI vision technologies offer considerable potential for increasing productivity and addressing labor shortages in the industry. A gradual approach, realistic expectations, and the willingness to collaborate with technology providers are crucial for success. Companies should focus on specific manual processes and carefully plan the integration of AI. With the right preparation and support, even complex variances in manufacturing can be reliably automated. Instead of waiting, industrial companies should actively start gaining their own experiences with AI in order to remain competitive and fully exploit the advantages of this advanced technology.

*Maximilian Mutschler is VP Sales at Micropsi Industries

(kip)