Get a grip! Not so easy if you are working with a robotic hand. Researchers from the University of Washington (UW) in Seattle are tackling the problem with the development of a sophisticated five-fingered robotic hand…one that can learn from its own experience without needing humans to tell it what to do. Vikash Kumar, a UW doctoral student in computer science and engineering, told OTW, “This robotic hand performs a set number of trials, and then goes for an improvement step based on the collective understanding obtained from the data of the performed trials.”
Regarding the design of their robotic hand, Kumar noted, “Our actuation system allows us to move the ShadowHand skeleton faster than a human hand (70 msec limit-to-limit movement, 30 msec overall reflex latency), generate sufficient forces (40N at each finger tendon, 125N at each wrist tendon), and achieve high compliance on the mechanism level (6 grams of external force at the fingertip displaces the finger when the system is powered.) This combination of speed, force and compliance is a prerequisite for dexterous manipulation, yet it has never before been achieved with a tendon-driven system, let alone a system with 24 degrees of freedom and 40 tendons.”
“There are a number of interesting robotic hands out there, but the fact that we have seldom seen them exploit their full dexterity in order to perform dynamics movements with free objects explains how challenging this problem is. Most real world applications resort to low dimensional custom designed (to fit the use-case) hands with stiff movements in order to sidestep the complexities of design, manufacturing and planning. If a robot butler or a fully functional prosthesis hand needs to become reality someday, they will likely need a dexterous manipulator capable of handling tasks of various nature and rigor. The goal of our research is to make progress towards this future by pushing the boundaries both in terms of design as well as planning/control of such devices.”
“Dexterous hand manipulation is one of the most complex types of biological movement, and has proven very difficult to replicate in robots. The usual approaches to robotic control—following pre-defined trajectories or planning online with reduced models—are both inapplicable. Dexterous manipulation is so sensitive to small variations in contact force and object location that it seems to require online planning. Here we are harmoniously bringing together concepts from the fields of Optimal Control and Machine Learning to design controllers that are capable of synthesizing dynamics dexterous maneuvers in real time for complicated high dimensional systems, while simultaneously learning from their own experience in order to improve their own abilities.”
Read More – Source: Sophisticated Robotic Hand Learns, | Orthopedics This Week