Open Collections

Multiagent manipulator control

University of British Columbia

The objective of this research is to define, specify, and implement a new, robust, and extensible manipulator control founded upon recent developments in multiagent robot control architectures.. Historically manipulator controllers serve within an idealized monolithic "sense-model-plan-act" (SMPA) control cycle that is both difficult and expensive to design for real time implementation. Recently, however, robotic systems have achieved remarkable performance through the combination of multiple, relatively simple, task specific controllers. These agents are arguably more reliable, robust, and extensible than SMPA architectures exhibiting similar performance. Furthermore, complex tasks have been achieved through multiagent teams, often exhibiting self organinzing or emergent behaviour. Despite these benefits and the growing popularity of these techniques, a formal model of agent and/or multiagent systems has not been proposed nor has any such architecture been applied to classical manipulation robotics. This thesis attempts to address these omissions through the analysis and application of multiagent design principles to manipulator control. After an introduction; to problems in real time supervisory robot control, an overview of manipulator and controller dynamics establishes a reference robot model. With this model as background, experimental high performance robot architectures arc examined and concepts common to these systems identified. Multiagent manipulator control strategies are then discussed and global goal distribution mechanism introduced. The design and implementation of a complementary multiprocess manipulator simulator is then described. With a global goal distribution definition, the design of a manipulator model-free global goal generator is discussed. Results from a multiagent manipulator control simulation are then presented and evaluated. The focus then turns to multiple global goal operation, discussing self organization, multiple global goals, and the impact of simultaneously active local and global goal systems on stability and arbitration. Results demonstrate multiple global and local goal operation including combinations of end effector trajectory tracking, joint failure, obstacle avoidance, joint centering, and joint limit avoidance. Finally, the significance of these results is discussed in the context of general multiagent control.

Thesis/Dissertation

application/pdf

2009-04-20

For non-commercial purposes only, such as research, private study and education. Additional conditions apply, see Terms of Use https://open.library.ubc.ca/terms_of_use.

http://hdl.handle.net/2429/7391

Mechanical Engineering

University of British Columbia

UBCV

DSpace