By Michael Jermann, Assistant Editor
A project intended to develop advanced prosthetics for military amputees may ultimately eliminate the need to put military personnel at risk to begin with, if engineers at Telefactor Robotics have their way.
“Robotic arms typically terminate in rigid two-fingered pinchers,” said Matt Kozlowski, vice-president of engineering at Telefactor. “We’re replacing that with a conformal finger system that can pick up an amorphous object with little or no additional cognitive load on the user.”
Robot arms that might ordinarily terminate in a standard end effector or a dedicated tool could be fitted with a conformal finger system that would in turn allow them to pick up a screwdriver, wrench, or welding fixture on command. This upgrade would not only create a multifunction arm, it could eventually enable that same robot arm to be programmed to pick up tools one by one at the appropriate time and perform a different sequence of movements with each.
“It’s a challenging task,” said Telefactor chief technology officer Stuart Harshbarger. “The human hand is a complex structure, with 20 or more degrees of freedom, depending on how you count, all packed into an extremely small package. It simply isn’t feasible from a size, weight, or cost perspective to replicate that architecture in a purely electromechanical system.”
For starters, the company stripped the design down to three digits: two “fingers” and an opposable “thumb.” Each digit has three joints; in addition, the thumb has the ability to adduct/abduct for opposition. Even this simplified structure nominally requires 10 degrees of freedom. Here, too, one motor per axis wasn’t feasible. Beyond size and weight considerations, every additional gear motor adds a point of failure. To reduce the complexity, the design team used a technique called passive actuation.
Passive actuation takes advantage of the interactive forces between the axis and environment, for example using linkages and ratcheting non-backdriveable sections to harvest potential energy and allowing part of the structure to move without active power. Kozlowski calls it under actuation.
The human finger includes three joints. Starting from the base of the finger at the palm, they are the metacarpal interphalangeal (MCP), the proximal interphalangeal (PIP), and the distal interphalangeal (DIP). Although humans can actuate two and often three of the joints independently, Kozlowski and his team realized that was unnecessary for a robotic hand. Instead, they placed the motor at the MCP and fit it with a spur gear. The gear interacts with an actuator through a device called a captured tendon mechanism, a cable “tendon” that connects to the PIP and the DIP. When the gear turns in one direction, the tendon tightens to flex the finger; when it reverses, the tendon drives in the opposite direction to actively extend the finger mechanism. That same gear also delivers torque directly to the MCP. “With a single under-actuated drive, we get effectively three degrees of freedom that are differential to each other,” said Kozlowski. By modifying the spring values and damping ratios at the various joints, they can tune a robotic hand to deliver the desired human-like motion.
The approach allows them to build a functional three-fingered robotic hand with 10 degrees of freedom, only four of which need to be directly actuated by motors. The four independent motors, all FAULHABER dc motors from MICROMO, operate in a coordinated fashion off of a single control bus. The resulting device is smaller, lighter, and consumes less power. Strategically locating the drives, coupled with the reduced overall mass lowers the effective inertia, improving responsiveness and speeding motion.
Another key part of their approach was to go with a modular design platform, a move intended to tame the complexity of the system. The integrated solution consists of a fully anthropomorphic system capable of the range of motion delivered by a human arm/body system. Each part presents different requirements for torque and range of motion. Using a different gear motor for each joint and degree of freedom would have required separate drives and controls, which would add size, weight, cost, and complexity. Instead, the team developed a modular design based on a well-defined and understood architecture. The same motor can be paired with different gear heads for reduction ratios ranging from 66:1 to 143:1.
“The fact that we’ve been able to parameterize our designs effectively with these transmission elements is really key because that means that we don’t have to size an entire new motor and then an entirely new controller and controller architecture,” said Kozlowski. “We just have to understand our speed and torque differential.” The team uses the same controller electronics for multiple motors; even the controls are multi-functional.