Adaptive Robotic Manipulator and Medical Device Design using Contextual Modeling and Numerical Optimization
Speaker: Frank Hammond, MIT
Department: Mechanical & Aerospace Engineering
Location: Engineering Quadrangle J223
Date/Time: Monday, October 21, 2013, 2:30 p.m. - 3:30 p.m.
The design of effective robotic manipulators and medical devices requires careful selection of mechanical system components, proper design of actuation topologies and morphological configurations, and solid understandings of the manipulation tasks and medical conditions that they are intended to address. Due to the complexity of the intended tasks and the immense size of the design spaces, developing these devices is non-trivial and can easily lead to solutions that are expensive, exorbitantly complex, and lacking the intended functionality and versatility. Our work demonstrates how the use of well-formulated numerical models, simulation methods, optimization frameworks, and the analysis of empirical data can mitigate mechanical complexity and elucidate salient design features to help us arrive at more functional and flexible robotic manipulation devices for a variety of applications.
In this talk I will discuss several examples of this design strategy, beginning with the use of heuristic, multi-objective fitness measures and evolutionary optimization algorithms to design adaptive, kinematically redundant robotic manipulators. This design novel approach produces manipulators that can resolve kinematic redundancy to achieve secondary goals such as energy efficiency and robustness in unstructured environments, which improve the economy and flexibility of automated industrial tasks, while preserving dexterity, precision, and design simplicity. Next, I will focus on the use of numerical grasp simulation and analysis techniques to optimize the design of underactuated robotic hands. In particular, I will discuss the derivation of non-anthropomorphic grasp synergies using an actuation topology reduction method, as well as the numerical exploration of the vast robotic hand design space to determine the viability of non-biomorphic hand morphologies which, despite their unnatural appearance, may prove more adept at grasping certain classes of objects than human hands. I will then describe the data-driven design of a dexterous robotic micromanipulation system for microsurgery, and how analyzing empirical data gathered from tracking the motion of surgical micromanipulation instruments can lead to the specification of more reliable and comprehensive system performance requirements. These requirements are used to optimize the kinematic and mechanical design of the robotic micromanipulation system to exceed manual micromanipulation capabilities, such as dexterity, precision, and repeatability, enabling new, more effective surgical intervention techniques. Finally, I will present recent work on the development of soft sensors and discuss future research topics and potential applications of this contextual modeling and design optimization strategy to human motion augmentation, adaptive robotic manipulation, and bioinspired mobile robots.