For humans and other mammals, sleep is more than just regeneration. It has a positive effect on the working memory and enhances complex cognitive functions such as decision-making and logical thinking. What if robots could make use of their inactive phases to improve their capabilities? Then they would be in a position to process their experiences and investigate future situations – to “dream” about them – and how they could best meet the anticipated challenges. This thought was the inspiration behind the RobDREAM project, which was named accordingly.
What are the advantages of a “dreaming robot”?
The two fundamental objectives of the project were easier generation and implementation of applications, and the enhancement of performance in mobile manipulation. To put it in colloquial terms: how can I get my robot up and running faster, and how can it continuously optimize how it performs its tasks. This is made possible by powerful algorithms. They ensure that autonomous navigation, perception, path planning and gripper technology can respond flexibly to changes in the environment.
That allows developers to compile an application as quickly as possible without having to bother with laboriously setting parameters and tuning the sequence. Even if the performance is initially not so great, the robot can at least start working. As it repeatedly performs its task, the robot gathers relevant data. In its inactive phases the robot can use these data to run the situations through a simulation and determine the best solution to apply in real operation. Thus the robot continuously optimizes its own performance.
What we have gained from the RobDREAM project
One central result is the mobile manipulator, a mobile robot platform with a robot arm. This has the capability to independently generate solutions to logistics tasks in control cabinet manufacture. In doing so, the robot responds to changes in its environment and can find the desired parts in the rack even if they are randomly distributed and not sorted – a useful capability especially in logistics.
In the project, KUKA developed methods to allow complex robot tasks to be programmed relatively easily. In addition it proved possible to expand the navigation interfaces in the course of RobDREAM. Object detection algorithms and methods for environmental modeling were also upgraded, now enabling the robot to distinguish very similar parts from each other and to determine their position.
Moreover, KUKA has developed procedures by which computationally intensive methods can be distributed across multiple computers working in parallel. In addition, the partners made significant progress in the automated optimization of parameters for navigation, perception and gripper algorithms.
From research to everyday industrial applications?
As a general rule, the results of a funded project are integrated into KUKA’s own systems to see whether these new technologies can be implemented in future solutions and products, and if so, how. “We can already state that the integrated complete system is a magnificent example of the capabilities of a mobile manipulator,” explains project manager Dr. Daniel Braun. “It shows interested employees and customers what can be achieved with KUKA systems. The application further advances the technologies developed by Corporate Research, particularly in relation to the stability, flexibility and usability of the results for KUKA.”
The scenario and robot application of RobDREAM are now being used further in the ProDok 4.0 project, which is concerned with automatic documentation. The integrated system and the development results in the individual technologies are being implemented in various other research projects. So KUKA robots can dream on and optimize their performance as we do when sleeping.
You can find out more about RobDREAM, ProDok 4.0 and other research projects here: https://www.kuka.com/de-de/technologien/konzernforschung/forschungskooperationen