Active safety in collaborative robots

Krogh, Simon Bjerre

4. term

Control and Automation, Master

2018

2018-06-07

97

Dette projekt undersøger, hvordan man kan opdage kollisioner i en robot med tre frihedsgrader (tre uafhængige bevægelser) ved hjælp af en dynamisk model. Først udvikles en PID-regulator, en udbredt feedback-metode til at holde robotten på den ønskede bane. Derefter afprøves et udvidet Kalman-filter, en metode der kombinerer model og støjfyldte målinger for at estimere skjulte tilstande, til kollisionsdetektion. Det udvidede Kalman-filter virkede ikke efter hensigten og blev erstattet af et lineært, stationært Kalman-filter, som kan estimere systemets tilstande. Selve kollisionsdetektionen bygger på at estimere ydre kræfter med en open-loop-tilgang (uden feedback). Denne tilgang viste sig utilstrækkelig, fordi kraftestimaterne var stærkt korreleret med styreinput. På grund af denne korrelation var det ikke muligt at afgøre, om der var sket en kollision. Rapporten afsluttes med en diskussion af problemerne og hvordan fremtidig forskning kan adressere dem.

This thesis explores how to detect collisions in a robotic system with three degrees of freedom (three independent movements) using a dynamic model. First, a PID controller, a common feedback method for keeping the robot on its desired path, is designed to control the system. Next, an extended Kalman filter—an estimator that combines a model with noisy measurements to infer hidden states—is tested for collision detection. The extended Kalman filter did not work as intended and was replaced with a linear, stationary Kalman filter, which can estimate the system states. The collision detection itself relies on estimating external forces using an open-loop approach (without feedback). This approach proved insufficient because the force estimates were highly correlated with the control inputs. Due to this correlation, it was not possible to tell whether a collision had occurred. The report ends with a discussion of these challenges and how future research might address them.

[This abstract was generated with the help of AI]

Collision Detection ; Collaboraive robots ; Faul detection ; Active safety

Download
View record in AAU Student Projects

Krogh, Simon Bjerre:

• Teleoperation of a surgical robot using force feedback (2016)