Adaptive gravity-balancing arm systems

Harshal Upadhyay, Purdue University

Abstract

Gravity balancing arms are passive weight support mechanisms that have been used to support human arms when weakened or otherwise in need of assistance. However, these systems could be greatly enhanced for everyday use if they can adapt for changing load mass or position. This thesis presents the development and preliminary testing of an adaptive system for gravity balancing arm devices that requires minimal user involvement and has low power and sensing requirements since it is built upon the system’s passive dynamics. It uses active control only to re-equilibrate the underlying passive system for changing conditions, then is turned off when not needed. Users can go about everyday tasks, and as a load mass or position for their task changes, they simply switch the system into an adaptation mode for either load mass or position, and the system takes care of the rest. The controller uses an indirect and low-power actuation method, adjusting the position of a key passive spring parameter (‘a’ value). The system requires only one sensor, an encoder, to measure the angle of the gravity balancing arm, which is used to indicate position of the gravity balancing arm. We use gain scheduling feedback control due to the nonlinearity of the gravity balancing arm system. Here, we primarily seek to demonstrate the feasibility of this novel system design. However, we also experimentally measure the adaptation response of the system for multiple load masses and two versions of the control gains (one for minimal damping to reduce energy cost, and one with increased damping effect to improve response times). We seek response times that are fast enough for the user to maintain task memory (2-4 seconds), but not significantly faster to keep power, weight, and actuator cost lower. We confirm that the system meets this objective by quantitatively measuring response times for each trial and providing a qualitative analysis of the system effectiveness based upon user-centered requirements from the field of user-interface design. Overall, we find that the system initiates physical adaptation changes fast enough to be perceived as continuous with the user’s task (less than 1 second), and can complete adaptation fast enough for users to maintain task memory (2-4 seconds) when load masses are less than 7.5 lbs. ^

Degree

M.S.M.E.

Advisors

Justin E. Seipel, Purdue University.

Subject Area

Mechanical engineering|Robotics

Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server
.

Share

COinS