A (0:02)

The human hand has 35 actuators, or muscles, for producing movement, 19 of which are inside the hand, responsible for fine movements like bringing the fingers together or pulling them apart. The other 16 are located outside the hand but extend into it via tendons. These provide power for gripping and lifting. Scientists can reproduce at least some of the hand's physical elements and capabilities. But the rub has been control. How do we convey to the prosthetic all the complicated movements behind a simple gesture like grasping? In the early 1960s, a Yugoslav scientist from the well known Mihailo Pupin Institute named Rocco Tomovich looked to biology for a solution. He leveraged his findings to create arguably the first autonomous robotic hand, certainly the first such five fingered robotic hand. And it was capable of precision tasks that might surprise you. In today's video, we profile a series of pioneering robotic prosthetics from Communist Yugoslavia. The Belgrade Hands. Tomovic was born in November 1919 in Baja, Hungary. In 1936, his family settled in Belgrade, where he did his secondary schooling. He then enrolled at the University of Belgrade to study electrical engineering. But those studies were interrupted by World War II. Tomovich joined the Yugoslav Communist Party and fought the Axis forces. He was arrested, sent to a forced labor camp and for some time worked in a mine. By the end of the war, he had risen to be a captain in the Yugoslav People's Army. After the war's end, he transferred back to the University of Belgrade to finish his studies. He earned his doctorate in the field of analog computers and then worked for a decade in Yugoslavia's Nuclear Research Institute in Vincia. In 1960, Tomovich left, perhaps due to Tito halting the Yugoslav nuclear weapons program. He worked at UCLA in 1961 as a visiting scholar and and then joined the Mihaylo Pupin Institute. There he helped produce one of the country's first computers, the vacuum tube and transistor base CER10. After that, he pivoted into work for developing prosthetics, reportedly due to a desire to help injured World War II veterans. People have been making hand prosthetics since at least the days of the Roman Republic. In 218 BC, the Roman general Marcus Sergius is said to have been fitted with an iron prosthetic that allowed him to hold a shield and fight battles against Hannibal of Carthage. In the 1500s, two iron prosthetics were built by Gutz von Berlickingen in Germany and Ambroise Paret in France. Driven by springs, levers and gears, users can initiate a grasp by the hands via exaggerated movements of the chest or arm. Such ways were generally how people controlled their prosthetic hands back then, which was impractical. But what other methods were available? In the 1910s, Ferdinand Sauerbruch, a famous German surgeon, created a natural looking prosthetic hand controlled by pins surgically inserted into the user's forearm. Ouch. That particular prosthetic hand was ultimately not used for cost reasons, plus all the infections and inflammation stemming from the necessary surgical preparations for the stump. But the hand introduced the idea of myoelectric control, meaning control using electric signals generated by the body's muscles, and this idea remained relevant. Significant work was done in the 1940s and 1950s in Germany and the Soviet Union. One rather unknown myoelectric Yugoslav pioneer was Ludovic Vodovnik of Slovenia, who partnered with Case Institute of Technology in Cleveland to pioneer assistive applications. I don't think he even has a Wikipedia page. But even these myoelectrically controlled systems were quite crude. You positioned and then commanded them to brainlessly execute a pre predetermined sequence to perform a simple, crude action. They were essentially glorified litter pickers. Can we imbue them with a little more smarts here? Tomovich argued that while the prosthetic hand's components and circuits had vastly improved after World War II, the mathematical theory to control them lagged behind. In a 1965 interview with the US newspaper, Thomas Tomovich said, we are not concerned with changing the hardware. Rather, we're working on introduction of new mathematics techniques in computer use. He sought to improve these theories by modeling after how biological systems work, taking in sensory inputs from the outside world and responding to them in a natural way. Regarding the artificial hand, he pointed to how existing systems relied on zero error. Such systems worked by monitoring their own state against an internal reference or command signal. The difference between the two was called error, and the goal was to maintain the smallest possible error. So if a robot hand's reference state says it should be at 30 degrees, but the actual hand sensors tell it it is at 28, then we seek to reduce that error of 2 degrees in subsequent movements. Zero error might be helpful for certain cases, like industrial robot arms, but is not suitable in life where you encounter new and different things all the time. Also, Tomovich noted that this was not how biological systems in the world functioned. So as an alternative, he proposed a reversal, a system that maximized error. He brought up the scenario of a hand clasping around something. And notice how hands, even those of a baby, achieve this by maximizing a feedback signal, in other words, the amount of skin area touching the target object. He realized that this simple theoretical model can be leveraged to flexibly handle diverse situations. Instead of telling the hand exactly where to go to clasp something, give it a few loose goals and conditions and let it figure it out on its own via an error maximization feedback loop. His experiment involved an electrically powered skeleton hand, same structure as a person's, with a special rubber glove on it. The glove is conductive, so pressure on the glove can produce an electric feedback signal that the system is incentivized to maximize. Once fully clasped, the signal stops rising and the hand stops applying pressure. Automatic feedback. Thus, the artificial gloved hand can close its fingers around a wide variety of differently sized objects without requiring any special programming for each object's individual size and circumstance. This flow of defining the end goal, which was to class the item but not rigidly defining the transition, states to achieve it and might seem rather obvious today, but it was a major leap from prior control theory regimes. Tomovich wrote in the paper's conclusion that he and his team intended to use these principles to build a full prosthetic hand and possibly a robot. Work on the first Belgrade hand began in 1963 and was presented in early 1964. Powered by an external power source, the Belgrade hand was about the same size as a normal human hand and mostly looked like one too. It was attached to the forearm and moved into action by a pressure sensor triggered by the user's biceps. It had five fingers, the first recorded robotic hand with that many. The four non thumb fingers mechanics operated like the natural finger, with three rigid sections connected with flexible joints. The thumb, however, worked differently. It can only rotate into one of two positions, one outside of the hand and the other inside the hand just opposite the middle finger. Building on the biology inspired feedback principles we spoke on earlier, the hand can grasp things of various sizes using one of two techniques. In the first, the thumb rotates to outside of the hand. When any of the finger pads touch an object, the four non thumb fingers then curl together around that object in a clenching movement. I'm somewhat reminded of those Venus fly traps. If the object is thick, like a book or something, then the finger's curling progress is impeded and the hand ends up in a position something akin to a hook. This can be useful for grabbing door handles, but if the object is small, then the fingers hit their limits first and then the thumb rotates inward from its original position to rest outside the clenched fingers, creating something like a fist. The second Major grasping movement that the hand can do was with its fingertips, kind of like an OK symbol with your hands. I imagine it might be, might be useful for picking up a small thing with precision. All of these movements were powered by a single geared motor located partially outside of the hand. That motor's force is then distributed using a complicated set of springs and levers. Pressure sensors in the fingers provided the feedback for the system to determine its movements. The big deal about the Belgrade hand was how it managed to take in and respond to feedback from the outside world. To do precision movements, you can make the argument that the thing itself had intelligence, taking on some of the mental burden from the user. And unlike another pioneering robotic hand, the MH1 produced at MIT by Heinrich Ernst in 1962, the Belgrade hand's intelligent capabilities were achieved without the use of a large digital computer attached to it. It seemed to me an analog tool reflecting Tomovich's great experience and in the space. A clinical evaluation of this first hand identified drawbacks. The first Belgrade Hans mechanical complexity made it difficult to manage, and its large external power system made it too heavy and bulky for practical use. But its capabilities were interesting enough to receive funding from the U.S. vocational Rehabilitation Administration in Washington. In the mid-1960s, the VRA was looking to develop an externally powered multifunctional hand arm system. So throughout 1966 and 1967, the robotics team at the Puppen Institute collaborated with the VRA and the U.S. national Academy of Sciences. The partnership allowed Americans to travel to Yugoslavia, a rare enough occurrence to sometimes make the local news. The second hand operated using the same control principles as the first, but with tweaks to the lever system, finger joints and thumb. The control logic was improved to offer the user more flexibility and predictability regarding how the hand grasped and opened. After completion, the hand was slated for demonstration alongside eight other candidates at the seventh Workshop on Upper Extremity Prosthetics Components in Santa Monica on 7-30-3169. Unfortunately, this second generation prosthetic was also too heavy to be fitted onto the tester and was not selected for use. A bit later, NASA considered using the hand as a possible end effector for some future space project, but this did not work out as well. But the experience evidently brought up the idea of the hand's use for robots. At the start of the 1980s, most robot hands were just simple two fingered grippers. But as the decade progressed, a number of things happened. Semiconductor microprocessors and components shrank in size while also improving in performance. Combined with increased investments from both government and industry, robot hands rapidly improved. Two of the most famous robot hands of this era were Ken Salisbury's three fingered hand and the Utah MIT Dextrous Hand. The first became an icon and an actual commercial, albeit somewhat expensive product. The hand's designer, Ken Salisbury, later went on to join Intuitive Surgical where he contributed to the famous Da Vinci surgical system. The second hand was a very advanced five finger hand that the National Science foundation funded other universities to purchase and study. This new environment brought out a renewed interest in the Belgrade Hand and In the late 1980s, Tomovich teamed up with longtime friend and collaborator George Becky at the University of Southern California to make a new version, the Belgrade USC Hand. With this robot hand, Tomovich and Becky wanted to demonstrate the validity of a new philosophy of control theory. The contemporary paradigm of robot locomotion was a complex top down system that receives input signals from sensors, processes those signals, recognizes patterns, infers actions in response and then plans out motions to execute. The two argued that this was too complicated and brittle for practical use, necessitating explicit mathematical models of trajectories and movement and whatnot. They instead argued for a philosophy of control based on stored reflex like responses triggered by learned sensory patterns. Such a philosophy is more akin to that of living animals and requires far less complex interplay. They called it artificial reflex control or arc. The Belgrade USC hand was intended to demonstrate that a robot hand can pick up and grasp items of varying size with just arc principles rather than complex top down instructions. In terms of physical changes, the new hand uses four separate motors for its non thumb fingers. For additional flexibility, the thumb was made to be fully articulated with two joints so it can rotate at an axis parallel to the wrist opposite the index, middle and ring fingers. This hand was designed to be mounted onto a robotic arm called a Puma 560 at the wrist. An IBM PC AT computer is added for higher level tasks, visually identifying the target object, determining a grasping strategy and and pre shaping and positioning the hand. Its max carrying capacity was £5, 5 or 6 Belgrade USC hands were eventually fabricated and sold to university research labs in the U.S. germany and Yugoslavia. But in the commercial space it was too complicated to compete against simple robot arm grippers. Becky tried to get funding for a third version but couldn't do it. He recalls later that a small company in California tried to commercialize the hand but failed to sell more than two or three, losing a lot of money in the process. Unfortunately, Yugoslavia entered turbulent times soon after the Belgrade USC hands presentation in 1990, Tomovich retreated from the public, passing away later in 2001 at the age of 81. Becky continued studying biologically inspired robot control throughout the 1990s. He passed last year at the age of 96. In a 2008 interview with Wired magazine, George Becky says that a robot hand today would not need five fingers. Hands like the Salisbury have largely proven that three fingers are good enough. But the Belgrade hand nevertheless stands as a pioneering achievement in robotics history, frequently mentioned in papers for its groundbreaking nature. And it and its creators are celebrated today as a sign of technological progress in old Yugoslavia. Even these remembrances largely skip over what I think was Tomovich's North Star, his biologically inspired theory of controlling robots via local reflex actions rather than complex top down systems. It's a radical idea, one that continues to fascinate people and spur on future research to this day. Alright everyone, that's it for tonight. Tonight was a short one. Have a good night. Sign up for the Patreon and I'll see you guys next time.