We can’t imagine industry anymore without robots, and they are now also becoming more common in the wider society. Their main purpose is to make life easier; machines can take over tasks from humans and often perform them more quickly, cheaply and better. Robots mow the lawn, vacuum- clean the house and assemble cars at the assembly line. But as yet they have limited flexibility and ease of use; a different robot is required for each task. We are currently seeing the trend that robots are becoming more complex and versatile. Our expectation is that they will be able to act independently in any situation. And that requires intelligence.
Article from Objective 22, 2014
Say robot and many people will think of movies featuring humanoid creatures that do exactly what they’re told to do. In industry, however, most robots consist of no more than a mechanical arm with a tool connected to it, which performs repetitive acts along an assembly line. Industrial robots excel in well-defined settings. They always carry out the same weld or tighten the same screw. They seem smart, because of their speed and accuracy. But that’s mainly the result of perfect mechanics – development of robots as mechanisms has advanced very far. But a welding robot can’t anticipate expanding metal or contamination of the weld. And if you want to ask a robot to do some other task, this requires a lot of programming and configuring. What you really want is a robot that needs only a few instructions and that anticipates changes independently.
Ease of use
We want smarter and more flexible robots. That primarily means making better driver software. Robot arms are like PCs: without software they can’t do anything. The development of intelligent robot arms is currently only in its infancy. Robots will undergo the same evolution as PCs, which evolved into the tablets and smartphones of our time that can be used for a huge range of tasks. In the future, robots will also be suitable for a wider range of tasks and will be much easier for humans to use and change. End users will be able to instruct robots in a very easy way. For instance through voice or image recognition, by showing the robot something or by demonstrating something. This would make an assembly line much easier to adapt to a new product or a new environment. You give the robot a simple set of instructions and then he is able to figure it out himself.
Intelligence and senses
This can only be achieved if robots can recognize exceptions and are able to deal with them. This self-reliance requires intelligence, but also senses, such as ‘eyes and ears’ to observe the surrounding world: sensors and cameras. A robot with eyes and intelligence will be able to independently see and recognize where something is and think of a way of picking it up. Or will be able to insert a pin into a passing object without knowing in advance when or at what speed it will be passing.
The price of sensors, cameras and computing power is no longer an obstacle. These elements are becoming increasingly cheap and increasingly widely available. However, universal robotic software is a challenge. The academic world has developed a Robot Operating System (ROS) to be able to exchange knowledge between research groups. Major manufacturers and market players have also begun working seriously on an industrial version: ROS Industrial. This is not a ready-made OS, but it is something that developers can use to develop further. It includes planning of movement, but you will have to add a feature yourself to register obstacles on the way with a camera and to act accordingly.
Rapid switching with smart robots
The trend in the manufacturing industry is to have smaller series and produce more tailor-made work. Take 3D printing, for instance: anyone can print their own unique product. Mass production has to think of an answer to this development. One possible answer might be to have more flexible production lines that can be easily switched to make more diverse products. This requires more intelligent robots that facilitate swift switching. They have to be usable more widely, have to be able to deal with complex tasks and to anticipate things that go wrong. And all this with software that can drive any robot, that is user-friendly and that can be used quickly by the user to program a robot for a new task.
A threat to your job?
Are smart robots going to be a threat to our jobs? No. There will of course be fewer people along the assembly line, but this development does create jobs in the surrounding sectors. In addition, robots can perform tasks in dangerous situations that humans can’t do. Developing smart robots does not mean that all assembly line robots have to become more intelligent. Simple conditioning for repetitive tasks suffices perfectly in situations where intelligence is not necessary. Smart robots can be deployed more widely where necessary, but there is a long road to go before they are ready to take over our work.
Robot plays Jenga
Technolution sees robotics as a major trend in the near future and wants to lead the way in offering intelligent robotics solutions. In order to acquire knowledge of and experience with steering robots, we are teaching a robot arm to play a game of Jenga. Jenga is a great means of solving practical problems. How do you build a orderly tower? How do you pick a block from a chaotic pile that keeps changing? Along the way we are acquiring universal knowledge that we can use in other projects.
The robot first piles up the identical wooden blocks of the game to make a tower. Then he follows the rules of the game: he removes blocks from the tower without toppling it. The movements that the robot has to make to do this are impossible to program in advance. The robot has to experience itself what is happening and has to adjust his actions accordingly. Optical feedback is an obvious solution, in which the robot software uses information from camera images to direct movement. The camera sees whether the robot is pushing out a block with the movement it is making and should therefore continue to push or whether the movement will topple the whole tower, so that he should choose a different strategy. See also the article ‘Vision in the loop’ in Objective 20.
No conditioned environment
An additional challenge is that the robot will have to be able to play this game in any random environment. The robot driver has to play the game as autonomously as possible and has to be able to anticipate changes or unknown situations in the environment. If the tower falls, he has to be able to rebuild it himself. This kind of situation where all the blocks are strewn about is always different than the last one. The system has to be able to solve that itself: picking up each block from the pile separately and rebuilding an orderly tower.