Modeling and Control of Robotics Manipulators

There is a strong trend within the robotics industry to build lighter and hence more mechanically flexible robots. As a result there is a need to use new, advanced motion control concepts where additional sensors, in addition to the conventional motor angular positions sensors, can be included. To make the result useful in an industrial setup the tuning of the controllers must be at least semi-automatic, for example, using model parameters identified from the real robot. Within this sub-project, the results from the past works in LINK-SIC sub-projects on identification and sensor informatics are used to provide improved control performance. The focus is mainly on the following areas:

• Iterative Learning Control based on sensor fusion of motor measurements and arm sensors. ILC enables the robot to reach an accuracy close to the repeatability level by compensating systematic errors when performing the same motion.
• Improved stiffness, robustness, and path accuracy by using additional sensors. Different control strategies and different sensor locations are examples of what will be covered.

In general, a feasible approach to the control problem is model based and provides design as well as tuning rules. One explicit approach is to use state space methods from the control community and combine them with methods from sensor fusion, e.g. Extended Kalman Filters, where it is possible to fuse information from low cost sensors, primarily accelerometers and IMUs, with information from the motor angular position measurements. The main challenge is to combine the control and sensor fusion for the highly non-linear, unstable plant, and produce robust high performance controllers. Another important fundamental problem that is considered is the fundamental limitation with regards to robustness and performance when no additional sensors are included in the control loop. This is important because it will still be the normal configuration for most industrial robots, motivated by the added cost and complexity when additional sensors must be included in the manipulator design.