The Belgrade Hand: A Landmark in Early Cybernetic Robotics and Anthropomorphic Prosthetics

Section I. Introduction: Establishing the Cybernetic Framework

The Intellectual Birth of Robotics and Cybernetics (Mid-20th Century)

The mid-20th century marked a profound shift in technological philosophy, moving from purely mechanical automation toward the integration of feedback and control systems, formalized under the field of cybernetics. Pioneered primarily by thinkers such as Norbert Wiener, cybernetics sought to understand control and communication in both the animal and the machine, viewing human physiological systems—the nervous system, musculature, and sensory feedback loops—as models for designing sophisticated machines. In the context of prosthetics, this intellectual movement identified the core challenge as bridging human biological intent (efferent signals from the nervous system) with the mechanical action of the artificial device (actuators), while simultaneously providing the user with meaningful sensory information (afferent signals) from the mechanical interface.   

The initial stages of robotics in the early 1960s were characterized by competing priorities: developing autonomous machine intelligence, perfecting industrial manipulators, and creating functional biomechanical replacements for human limbs. The successful realization of an effective artificial limb required an unprecedented convergence of disciplines, including electrical engineering, anatomy, materials science, and control theory.

The Unveiling of the Belgrade Hand (1963/1964)

Amidst this global scientific ferment, a defining breakthrough occurred in Yugoslavia. The first model of a multifunctional externally powered hand prosthesis, known internationally as the ‘Belgrade Hand Prosthesis,’ was developed at the Institute 'Mihailo Pupin' in Belgrade. The theoretical foundation for the anthropomorphic device was laid in 1962 by Rajko Tomović and G. Boni. The subsequent physical device, which included five fully functional fingers, is officially credited as the world's first externally powered artificial hand of its kind, with its unveiling generally dated to 1963. A subsequent version was introduced in 1964. This timing places the Belgrade Hand precisely at the critical period in robotics history, bridging the first rudimentary bio-signal controlled devices and the sophisticated artificial intelligence manipulators developed shortly thereafter (e.g., Minsky’s work in 1965).   

The development occurred within a unique geopolitical environment. Belgrade, as the capital of non-aligned Yugoslavia, maintained scientific exchange across political divides. Professor Rajko Tomović was known for his comprehensive international communication with colleagues in the Soviet Union, European nations, and North America. This unique position allowed the research team to rapidly synthesize pioneering technological knowledge from disparate global sources. For example, myoelectric control, which was being aggressively developed in the Soviet sphere (e.g., the Russian Hand ), was combined with advanced cybernetic and systems theory prevalent in Western research (e.g., US robotics principles ). The success of the Belgrade Hand was thus, in part, a testament to this intellectual synthesis, unifying diverse technological streams into a superior and functionally complex prosthetic design.   

Structure of the Report

The ensuing analysis details the academic and institutional foundations that gave rise to this innovation, investigates the complex mechanical architecture, including the critical use of underactuation, dissects the revolutionary dual-layer cybernetic control system, and concludes by placing the Belgrade Hand within the global context of the 1960s robotics era, assessing its enduring influence on modern biomechatronics.

Section II. The Belgrade Cybernetic School: Pioneers, Institutions, and Philosophy

Profiles of the Visionaries

The development of the Belgrade Hand was driven by key scientific figures associated with the University of Belgrade and the Mihajlo Pupin Institute. Central to this effort was Professor Rajko Tomović (1919–2001), a Serbian and Yugoslav scientist whose career was marked by deep involvement across multiple pioneering fields, including Physics, Robotics, Medical Information Technology, Biomedical Engineering, and Rehabilitation Engineering. Tomović earned his doctorate in Technical Sciences from the Academy of Sciences of Serbia in 1952, establishing the interdisciplinary foundation necessary for the project.   

Tomović was joined by Professor Milan Rakić from the Faculty of Electrical Engineering at the University of Belgrade. The conceptual groundwork, which specifically addressed strategies for biological movement and anthropomorphic design, was initially documented in 1962 by Tomović and Boni. This collaboration harnessed deep expertise in systems theory, control mechanisms, and applied electrical engineering.   

The Institutional Nexus: The Mihajlo Pupin Institute

The research was carried out at the Institute "Mihajlo Pupin," an institution mandated for advanced applied engineering and automation. The environment fostered high-level, complex scientific work necessary for such an interdisciplinary endeavor.   

The institutional legacy of the Mihajlo Pupin Institute extends beyond manipulation to locomotion, demonstrating a pervasive focus on the mathematical control of complex anthropomorphic systems. The institute is also recognized as the home of the seminal Zero-Moment Point (ZMP) concept. The ZMP principle, developed to solve the problem of dynamic stability in bipedal locomotion, shares a fundamental, philosophical link with the control issues addressed by the Belgrade Hand. Both grasping (addressed by the Hand) and walking (addressed by ZMP) are complex challenges involving dynamic stability in multi-jointed, bio-inspired robotic systems. The concurrent development of these two landmark concepts—one concerning lower-extremity balance and the other upper-extremity manipulation—affirms a core competency within the Institute in the 1960s: the advanced mathematical modeling and engineered control of anthropomorphic robotics.   

The Cybernetic Philosophy: Anthropomorphism as a Design Constraint

The team’s approach was rooted in a cybernetic philosophy that placed high value on anthropomorphism, defined as replicating human-like structure and movement fidelity. Unlike previous prosthetic attempts that focused solely on basic gripping capability, the Belgrade Hand was conceived based on biomechanical modeling of the natural human hand.   

A primary design constraint was the necessity to enable complex actions such as thumb opposition and human-like grasps, ensuring appropriate force distributions among the fingers. This commitment to biofidelity dictated the technological complexity. Achieving this level of functional accuracy required an unprecedented integration of anatomical knowledge, advanced electronics for signal acquisition (EMG), intricate control theory (Finite-State Machine logic), and innovative mechanical engineering (underactuation). This interdisciplinary necessity established the Belgrade Hand project as a precursor to modern biomechatronics, where expertise must span human biology and sophisticated engineered solutions simultaneously.   

Section III. Architectural and Kinematic Innovation: Mastering Dexterity

The Mechanical Blueprint: Five-Finger Design

The most immediate distinguishing feature of the Belgrade Hand was its comprehensive mechanical architecture. It was specifically credited with being the first artificial hand to feature five functional fingers, representing a significant advancement in potential dexterity and natural motion compared to contemporary devices. This design was paramount in achieving human-like grasps, as true dexterity is impossible without the capacity for specialized thumb opposition.   

Degrees of Freedom (DoFs) and the Underactuation Strategy

The challenge inherent in replicating the human hand is the extraordinary number of mechanical degrees of freedom (DoFs) required for natural movement. The human hand utilizes many joints, demanding correspondingly complex control signals. In the 1960s, external power systems and myoelectric controls could not provide independent control channels for every joint.

The ingenious solution adopted by Tomović's team was the incorporation of underactuated mechanisms. Underactuation is a design philosophy where the number of actively controlled actuators is fewer than the total number of mechanical degrees of freedom (DoFs). This strategy allows the hand to passively adapt its complex geometry to the shape of the object being grasped, distributing forces appropriately without requiring explicit, active control input for every single joint.   

Quantifying the design complexity reveals the scale of this engineering feat. The system modeled the limb with 12 internal (controlled) degrees of freedom, which correspond to the joints capable of exerting torques. Additionally, there were 6 external (uncontrolled) degrees of freedom related to the position and orientation of the trunk. This revolutionary ratio—achieving functionality across 12 DoFs using a minimized set of control signals—demonstrated how mechanical optimization could substitute for computational power and control input abundance, a principle now foundational in modern manipulation and robotics.   

The decision to utilize underactuation was not merely a design preference but a necessary engineering response to the limitations of early electromyographic (EMG) control technology. Even with contemporary sophisticated techniques, residual limb musculature could typically provide inputs suitable for a maximum of two degrees of freedom using conventional on-off control. To achieve a fully articulated, functional, multi-finger hand with 12 internal DoFs, the designers had to shift the burden of complexity from the constrained electronic controller to the mechanical linkage, proving that high functional complexity could be attained through optimized mechanical compliance.   

Section IV. Control Architecture: The Myoelectric and Cybernetic System

Principles of Electromyographic (EMG) Control

The Belgrade Hand was externally powered and used myoelectric (microelectric) control. This technology relies on detecting electromyographic (EMG) signals—the small electrical impulses generated by muscle contraction in the residual limb—to serve as the efferent control inputs for the prosthesis.   

Early myoelectric methods were constrained. Simple on-off myoelectric control was generally suitable for a maximum of two DoFs. Direct myoelectric control introduced proportional capability using independent EMG sites for individual movement control. The Belgrade Hand, with its 12 controlled DoFs, required a sophisticated control structure capable of translating sparse, noisy biological signals into a complex repertoire of actions.   

The Dual-Layer Cybernetic Control System

The control architecture was a crucial element of the hand’s cybernetic design, structured to manage complex movement patterns and maintain stability based on limited user input. The system was composed of two main operational levels :   

1. High-Level Control: This layer was dedicated to interpreting the subject's intent and subsequently launching appropriate action patterns. Its function was crucial for mapping the sparse EMG inputs onto the 12 mechanical DoFs, essentially creating an algorithmic representation of user desire.   

2. Low-Level Control: This layer assumed responsibility for ensuring grasp stability once contact with an object was established. This automatic stabilization function provided the necessary operational reliability, preventing accidental dropping or crushing of grasped objects.   

Sensory Feedback and the Finite-State Machine (FSM)

A revolutionary aspect of the Belgrade Hand was its inclusion of sensory feedback, a feature deemed essential for a truly closed-loop cybernetic system. This sensory system, providing afferent signals back to the control system, consisted of two integral parts :   

1. Proprioceptive Subsystem: This provided information regarding the hand's internal state, monitoring kinematics (position, joint angles) and the internal forces generated within the hand's transmission mechanism.

2. Exteroceptive Subsystem: This subsystem monitored and measured the interaction forces occurring between the hand and the grasped object, as well as the interaction between the object and the external environment.   

Hand operation was organized as a Finite-State Machine (FSM). The transitions between different operational states (e.g., transitioning from a "reaching" state to a "grasping" state, or from a "grasping" state to an "exploring" state) were identified and triggered by the critical events detected by the sensory system. For instance, object contact or a change in load would trigger a transition. This design demonstrated contextual control—the hand's functional output for a given myoelectric signal was dependent on its current sensory state. This high-level control that actively sought to "interpret the subject's intention" established the Belgrade Hand as a crucial precursor to modern intent estimation algorithms, such as pattern recognition and regression modeling used today to realize simultaneous, proportional, and independent multi-DoF control.   

The integration of proprioceptive and exteroceptive feedback with an FSM defines the Belgrade Hand as one of the first complete, functioning closed-loop systems in bio-signal controlled robotics.

Table 1: Core Technical Features of the Belgrade Hand (1963/1964)

Feature Category

Detail/Specification

Significance

Development Date

1963 (Concept 1962) / Prototype 1964

Landmark date for multifunctional prosthetics [2, 4, 5]

Key Inventors

Rajko Tomović, Milan Rakić, (and Boni on theory)

Founders of cybernetic limb control [2, 3]

Control Mechanism

Myoelectric (Microelectric) Control

Utilized EMG bio-signals 

Control Strategy

High-Level/Low-Level Control, Finite-State Machine (FSM)

Interpreted user intent based on sensory feedback 

Mechanical Design

Anthropomorphic (Five Fingers)

Enabled human-like dexterity and thumb opposition [1, 4]

Internal Degrees of Freedom (DoFs)

12 (Controlled)

Measure of potential mechanical complexity 

External Degrees of Freedom (DoFs)

6 (Uncontrolled, e.g., position/orientation)

Defines the kinematic envelope 

Key Mechanical Principle

Underactuated Mechanisms

Allowed passive adaptation to object shapes 

Sensory Feedback Channels

Proprioceptive and Exteroceptive

Established a crucial closed-loop control system 

Section V. Comparative Analysis: The Belgrade Hand in the Global Pantheon (1960–1965)

The prominence of the Belgrade Hand can only be fully appreciated when compared against its immediate contemporaries during the crucial period of 1960 to 1965, a time of rapid advancement in both prosthetics and fundamental robotics research.

The Immediate Precedent: The Russian Myoelectric Hand (1960)

The first significant milestone in myoelectric prosthetics occurred in the Soviet Union. Russian scientist Alexander Kobrinski unveiled the first clinically successful myoelectric prosthesis in 1960. This device pioneered the use of transistors to reduce bulk and enable portability, placing the batteries and electronics on a belt connected by wires to the hand.   

While groundbreaking in its clinical application, the Russian Hand faced limitations. It was heavy, its movement was slow, pinch force was weak, and the wired connections were susceptible to damage and electrical interference. Fundamentally, the Russian Hand’s priority was validating the control method and achieving basic clinical utility (e.g., a simple grasp). In contrast, the Belgrade Hand, developed shortly thereafter, pioneered the multifunctional, anthropomorphic design enabled by myoelectric control, moving the goalpost from basic utility to complex dexterity.   

The American Counterpart: Minsky’s MA-3 Arm (1965)

In the United States, research often focused on autonomous machine intelligence. Marvin Minsky and Seymour Papert at MIT developed the MA-3 Arm (also known as Minsky's Tentacle Arm) starting around 1965. The purpose of this device was rooted in artificial intelligence (AI) research: creating a computer system that could autonomously "see" and "manipulate" objects, independent of direct human control. Minsky’s work sought to model the complexity of the human mind through simple processes called "agents".   

Technically, the MA-3 Arm diverged significantly from the Belgrade Hand. The MA-3 primarily employed pneumatic actuators, chosen for their ability to provide smooth, fluid, compliant, and lightweight motion necessary for experimental manipulation. The MA-3 was not designed for bio-signal integration, but rather for algorithmic autonomy.   

The MIT Museum currently displays the Belgrade Hand alongside the MA-3 Arm, underscoring the convergence of global efforts in the 1960s. This exhibit captures the defining philosophical debate of early robotics: the Belgrade Hand represented the approach of human integration (biomechanics and bio-signal control), while the MA-3 Arm represented the approach of machine intelligence (autonomy and algorithmic control).   

Collaborative Evolution: The USC Belgrade Hand

The international significance of the Belgrade Hand was cemented by subsequent transatlantic collaboration. Professor Tomović partnered with roboticist George A. Bekey at the University of Southern California (USC), resulting in the development of the "USC Belgrade hand". This collaboration confirmed the device's technical credibility and established it as a landmark in the global history of robotic hands, successfully transferring sophisticated control and design knowledge across continents.   

Table 2: Comparative Analysis of Early 1960s Robotics and Prosthetics

Device Name

Year (Approx.)

Primary Function/Focus

Key Technical Innovation

Anthropomorphism Level

Historical Claim

Russian Myoelectric Hand (Kobrinski)

1960

Clinical Function (Grip)

Transistors enabled portability/myoelectric control

Low (Limited function)

First clinically significant myoelectric prosthesis 

Belgrade Hand (Tomović/Rakić)

1963

Multifunctional Grasping/Dexterity

Underactuation, FSM, Sensory Feedback

High (Five fingers, biomechanical modeling)

First externally powered multifunctional hand prosthesis 

MA-3 Arm (Minsky)

1965

AI Manipulation/Machine Vision

Pneumatic Actuation, AI algorithms

Moderate (General flexible arm)

Pioneer in non-anthropomorphic robotic manipulation [5, 7]

The comparative analysis demonstrates that the Belgrade Hand occupied a unique niche. While Kobrinski prioritized rapid clinical adoption and Minsky prioritized machine autonomy, Tomović rigorously prioritized biofidelity and complex control. This insistence on high anthropomorphism—the five-fingered, dexterous design—forced the Belgrade team to invent mechanical and control solutions (underactuation and FSM logic) that were not required by the other devices, fundamentally advancing the state-of-the-art in human-machine interface design.

Section VI. The Enduring Legacy in Modern Biomechatronics

Foundational Influence on Modern Prosthetics

The conceptual and mechanical breakthroughs introduced by the Belgrade Hand established the essential blueprint for all subsequent high-end functional hand prosthetics. The requirement for a five-fingered, multi-functional design that allows for intricate grasping maneuvers is now the mandatory standard for advanced modern prostheses. The scientific investigation initiated by Tomović and Boni into anthropomorphic devices provided the seminal work that investigators followed, advancing biomimetic devices for both upper and lower extremity applications.   

The Principle of Underactuation and Mechanical Intelligence

Perhaps the most significant mechanical contribution was the widespread adoption of underactuation. The hand demonstrated the vital principle that functional complexity can be achieved by optimizing passive mechanical compliance, thereby dramatically reducing the complexity required of the active control electronics. Achieving high DoF functionality with low DoF control remains central to contemporary designs, particularly in adaptive grasping robots, where soft and compliant mechanics allow systems to robustly handle diverse objects with minimal explicit programming.   

Philosophical Impact: Redefining Human-Machine Interface

Tomović’s team redefined the prosthesis not merely as a replacement tool, but as a critical element within a human-machine control feedback loop, realizing the early ideals of cybernetics. The focus on sensory feedback (proprioceptive and exteroceptive) established the necessity of closed-loop operation for intuitive prosthetic control.   

This philosophical shift paved the way for advanced bio-integration methods developed decades later. For example, modern techniques such as targeted muscle reinnervation (TMR), which reroute amputated nerves to residual muscle sites to create more independent control channels and provide sensory feedback via the reinnervated skin , directly fulfill the functional goals articulated by the Belgrade Cybernetic School in the 1960s.   

The Evolution of Control Systems

The fundamental problem identified by the Belgrade team—the disparity between the richness of biological movement (12 DoFs) and the poverty of accessible biological control signals (limited EMG inputs)—is the core challenge driving modern biomechatronics research. Early systems were limited to simple on/off control. The Belgrade Hand’s Finite-State Machine logic represented the first systemic attempt to move beyond simple signal amplification to signal interpretation, creating a rule-based algorithm to estimate user intent.   

This lineage leads directly to contemporary methods that employ machine learning, such as pattern recognition and regression modeling, to extract features from myoelectric signals and estimate simultaneous, independent, and proportional control for multiple joints. Every advancement in simultaneous control today is built upon the foundational cybernetic realization that the artificial limb must respond not merely to a detected signal, but to an estimated contextual intent.   

Conclusion: The Belgrade Hand as a Rosetta Stone

The Belgrade Hand, developed in 1963/1964 by Tomović and Rakić, represents a critical nexus in technological history. It successfully merged the emerging fields of cybernetics, robotics, and rehabilitation engineering, creating the world's first externally powered, multifunctional, anthropomorphic hand prosthesis. Its innovative use of underactuation compensated for the limitations of early myoelectric control, while its dual-layer control structure and integrated sensory feedback established the requirement for a closed-loop system capable of interpreting contextual user intent.

The legacy of the Belgrade Hand extends far beyond its immediate clinical application, providing the foundational principles—biomechanical fidelity, mechanical intelligence via underactuation, and sophisticated cybernetic control—that continue to drive research in high-performance biomimetic robotics and advanced prosthetic limb design worldwide.