Recent robotics-automated manufacturing process requires not only assigning tasks but also planning motions among multiple robot machines. Until recent days, this kind of works have been done by human experts. In general, it takes a lot of trial-and-errors and verification processes to manually create a program that accomplish the objective tasks while avoiding collision and interference from each other robot machine.
We use domain knowledge to create an objective function that satisfies the constraints of the actual process and also employ various ML and optimization techniques to automatically assign each task to the robot. Path planning is carried out in accordance with the assigned task in order to create a fast and safe robot program. Additionally the algorithms implemented as modules support parallel executions in order to improve the speed and efficiency of route search.
The automatically generated robot program effectively performs the objective processes in a assembly line comprised of multiple robots. Work that used to take months is now completed in a week.
Related Use Cases
Control and optimization
Chip Placement with RL
Developing chip simulation environments and placement algorithms to train RL models that optimize element placement.
In place of an HVAC simulator, the trained dynamics model is used to predict changes in the HVAC's internal state, and it serves as the basis for development of an efficient air conditioning control model.