A compliant control law for industrial, dual-arm manipulators

Date

2013-05

Journal Title

Journal ISSN

Volume Title

Publisher

Abstract

Many of the first robots ever built, decades even before the first industrial robots, were humanoids. It seems like researchers have always sought to imitate the human form with their robots, and with good reason. Humans are incredibly flexible; they can perform a huge variety of tasks, from locomotion over rough terrain, to delicate assembly, to heavy lifting. A human’s second arm allows him to lift twice as much weight. His workspace is approximately doubled, and he can perform a broader variety of tasks as items are passed back and forth between hands. We sought to impart some of that same functionality to a strong, rigid, dual-arm robot. Specifically, we developed a control law that allows two robot arms to lift and manipulate an object in cooperation. As opposed to the prior art, our control law is tailored for industrial robots. These robots do not usually allow torque control and their control frequency is generally 60 Hz. Through the use of fuzzy logic, the control law is quite robust at 60 Hz control rates. Its simple structure reduces the computational cost of the algorithm by approximately 75% over Jacobian-based methods. Stability is proven and the controller parameters can be adjusted to handle perturbances of arbitrary magnitude. Since the robots behave as an admittance, torque control is not required. Several experiments were conducted to benchmark and validate the performance of this control law. The controller is able to maintain a clamp force within ± 4N despite a wide variation in trajectory and control frequency. This fine level of force control makes the controller suitable for delicate tasks. The conclusion suggests several extensions that would make this control law more useful. For example, adaptive control would improve the performance. A position feedback controller should be cascaded so that the robot arms’ tracking accuracy is improved. Many tasks (such as co-robotics) require external compliance, and we show how external compliance could easily be incorporated.

Description

text

Citation