Robotics

robot
We designed control algorithms for a bipedal balancing robot with only pegs for feet

Physical Modeling

In order for a robot to act usefully, it must have a detailed physical model of its construction. In some robots, this contains not only the lengths, angles, gear ratios, etc., but also its distribution of mass and the mass of whatever it is holding or carrying. This is particularly important for robots that must perform a task as quickly as possible. A robot carrying a heavy load to a destination must know how heavy the load is so that it can begin stopping early enough to not overshoot its target.

Control Theory

There is a lot of theory on how to control a robot to accomplish a particular task well. For simpler systems, the physical model and the defined task can be used to analytically derive a controller that will do the job. If the robot has an unknown mass (such as when it is picking up objects of unknown weight), then it may need an adaptive controller. In complicated problems, such as when the robot is when the robot is juggling devil sticks (this has actually been done by Schaal and Atkeson, although it would be nice to see a bit more complex maneuvers), it becomes useful to have the robot actually learn the problem, instead of just adapting to a changing environment. This then becomes reinforcement learning problem.

Factory Robot Challenges

Modern manufacturing plants rely more and more on robotic arms and systems for process automation. Often, systems supplied by vendors may not perform as expected due to insufficient characterization by the vendor, or incomplete understanding of the the peculiarities of robotic devices by the customer. Issues of accuracy, precision, repeatability of motion, and path selection may limit the usefulness of the device, or make it more difficult to use than expected. The equipment may not interface directly to existing equipment, thus requiring addition hardware and software.

Characterization studies of the arm, coupled with analysis of specific tasks, can lead to significant improvements in robotic arm capabilities for a given task. If you’re considering a robotic system, or need help with your current system, please contact us today!