Recently Foundation Models (FM) has attracted a lot of attention. These FMs most likely will have significant impact on how we will use robots in assembly and disassembly lines. For those who do not know FMs, these are models that were trained on internet-scale data and aim to provide very powerful retrieval capabilities for the algorithm with a natural language capabilities on the other.
As we see it, the most influence of FMs on Robotics will be in the following aspects:
- The English language will become the language of robotics. Currently, when we need to describe to a robot what it should do, we need a degree in Mechanical or Industrial Engineering, and we need use some software that describes the robot what to do. In general, this approach is referred as automation and it involves for the end use a lot of engineering. On the contrary, using FMs, and specifically, the using of language, images, movies, voice, etc. for operating a robot and describing what we would like the robot to do, will lower the operating barrier for many small and mid-size businesses and enable not only heavy enterprises to harness the power of robotics.
- Enable new Skills to robots. Many operations today are still performed by human operators. The reason for that varies from product and industry, but common lines shared between these verticals. For example, in the decades of automation, operations that involve tactile sensing or tactile assembly (such as insertion of shafts into holes), the tactile feedback is crucial. First, in many cases it may reduce the tear-and-wear of both the robots and the produced goods. Second, if the tolerance of the parts or the accuracy of bringing the parts to the right location is too small, the insertion may fail partially or completely. AI in this case may solve long standing challenges that could not solved with classical automation algorithms.
- Cost reduction through labour costs reduction. Many limited series operations today still require manual labour. Training robots to do operations that exclusively done by humans, or operations where the labour cost is smaller than bringing automation experts, may now become attractive.