Laser-Focused on Robot Accuracy

No doubt you have heard the jokes about how expensive boating is, like the one that quips “boat” is actually an acronym that stands for Break Out Another Thousand. But that is nothing compared to robotics, where costs are amplified by an order of magnitude or more. Because of these hefty expenses, organizations with industrial robots have to work hard to stretch the useful lives of their systems. That often means retrofitting their robots with additional hardware so that they can take on new tasks, rather than purchasing entirely new equipment.

For all sorts of use cases, it is necessary to give robots a greater level of precision in their movements. This can be particularly important to applications like assembly, where tiny miscalculations can result in the production of defective products. A common way of upgrading an existing robotic system is to fit it with a new active gripper with higher motion resolution. A relatively quick upgrade like this can minimize downtime, maintenance expenses, and production costs.

However, the accuracy of a robot upgraded in this way is typically not as good as an entirely new, purpose-built system. But a group of engineers at the University of Nottingham in the UK has come up with a way to enhance the accuracy of modified robots — and it involves lasers! Using their method, positional accuracy of robots was shown to improve by as much as 82%.

In their research, the team started with a standard six-degrees-of-freedom (DOF) industrial robot that uses revolute joints. To improve its motion resolution, they added a high-precision linear stage to the end effector, effectively transforming the robot into a seven-DoF “6RP” system, with the “P” indicating the inclusion of a prismatic (sliding) joint.

To achieve precision control over this modified robot, the engineers implemented a closed-loop feedback system using a high-end industrial laser tracker. This device provides real-time 3D positional data of the robot’s end effector, allowing the system to continually correct for errors.

To interpret and respond to this data, the team trained a multilayer perceptron neural network to model the robot’s forward kinematics — that is, how each joint’s position affects the final location of the end effector. This model enables the control system to use gradient descent, an optimization technique, to fine-tune the robot’s movements on the fly.

Testing on ISO-standard motion trajectories revealed a dramatic boost in positional accuracy. Over 30 test runs at five different points, robots retrofitted with the laser-assisted control system demonstrated an average 82% improvement in positional accuracy compared to their original configurations.

This method has the potential to offer cost savings across applications like precision assembly, inspection, and additive manufacturing. Instead of discarding or replacing older equipment, manufacturers can now supercharge existing robots to meet modern demands for precision.