Robotic arms learn delicate touch from bionic prosthetic hands used by real people
A groundbreaking partnership between industrial robotics leader ABB and California-based bionics firm PSYONIC aims to transform how robots handle fragile and irregular objects. The collaboration centers on using real-world touch and movement data collected from human prosthetic users to train robotic arms, potentially bridging a long-standing gap in automation capabilities. The approach marks a shift from laboratory-only robot training to learning from actual human interaction with everyday objects.
The initiative combines PSYONIC’s Ability Hand, originally designed as an advanced prosthetic device, with ABB’s GoFa collaborative robot. The bionic hand features multi-articulating fingers, pressure sensors, vibration feedback and flexible mechanics that allow it to conform to objects of varying shapes and textures. This technology captures how humans naturally adjust grip strength and finger position when handling different items, from delicate eggs to sturdy tools.
Why human grip remains difficult for industrial robots
Industrial robots excel at repetitive tasks involving speed, precision and endurance that would quickly fatigue human workers. However, they struggle significantly when confronted with objects requiring subtle touch adjustments. A coffee cup demands different grip pressure than a screwdriver. A soft package needs gentler handling than a metal component. Most humans make these adjustments instinctively, without conscious thought.
For robots, this adaptive capability represents a major engineering challenge. Too much pressure can damage fragile items or packaging. Insufficient grip strength causes objects to slip and fall. Even minor changes in object angle or position on a conveyor belt can disrupt automated handling processes. Marc Segura, president of ABB Robotics, identified human dexterity as “one of the most difficult things to replicate in industrial-grade robotics.” The current collaboration seeks to close this persistent performance gap.
How bionic hand data trains robotic systems
The PSYONIC Ability Hand was engineered to help amputees perform daily activities with greater ease and naturalness. It employs myoelectric control, which detects electrical signals from remaining muscles to control finger movements. Touch sensors throughout the device detect pressure during grips, while vibration feedback communicates tactile information back to the user. This lightweight design incorporates compliant mechanics that allow fingers to adapt to object contours.
When prosthetic users interact with objects in real-world situations, the Ability Hand generates detailed datasets about movement patterns, contact points and grip force variations. Dr. Aadeel Akhtar, founder and CEO of PSYONIC, characterized dexterous manipulation as “a data challenge as much as a hardware challenge.” The company believes that training data derived from natural human use provides more realistic and diverse information than demonstrations performed exclusively in controlled laboratory settings.
ABB’s GoFa collaborative robot provides the industrial testing platform for implementing these learning algorithms. The cobot offers the precision and repeatability necessary to validate whether movement and grip strategies learned from prosthetic use translate effectively to factory and warehouse applications. This integration represents what ABB calls Autonomous Versatile Robotics, or AVR, which envisions robots capable of sensing, reasoning and moving with precision in dynamic, changing environments.
Leia Também
Resultado Quina de São João revela ganhadores com prêmio de R$ 260 milhões em sorteio especial realizado neste domingo
Sorteio especial de R$ 260 milhões agita o domingo com prêmio que obrigatoriamente sairá hoje
Motivos que levam a Rockstar a ocultar novidades de Grand Theft Auto 6 no Summer Game Fest
Industries that could benefit from touch-enabled automation
The technology holds promise across multiple sectors where delicate or variable handling currently limits automation potential. Key industries identified by both companies include:
- Automotive manufacturing, particularly for soft interior components and complex assembly tasks
- Aerospace production, where precision handling of lightweight composite materials is critical
- Packaging operations requiring adaptation to products of varying sizes and fragility
- Logistics and fulfillment centers managing diverse product inventories
- Life sciences and medical device manufacturing demanding sterile, gentle handling
In these sectors, robots already perform substantial work, but delicate or inconsistent handling requirements often necessitate human intervention. The International Federation of Robotics has noted that advanced gripping capabilities combined with digital integration could reduce engineering setup time by up to 30 percent. Faster deployment and greater flexibility would allow companies to automate tasks that currently require extensive custom programming and adjustment.
Balancing automation advances with workforce concerns
Robots capable of handling repetitive or ergonomically challenging tasks could reduce physical strain on workers, particularly for jobs involving the same motion repeated thousands of times daily. This capability might prevent workplace injuries and improve long-term worker health outcomes. However, more versatile robots also raise questions about employment in sectors where variable handling tasks currently require human workers.
The most beneficial implementation would deploy robots for physically demanding repetitive work while humans focus on oversight, quality control, machine setup and higher-skill responsibilities. This approach could enhance productivity without simply replacing workers wholesale. The balance between automation benefits and labor market impacts will likely depend on how companies choose to integrate the technology and whether they invest in workforce training for new roles.
Real-world data approach changes robot training methods
The ABB Robotics and PSYONIC partnership represents a departure from traditional robot training methodologies. Rather than programming robots exclusively through controlled demonstrations or purely simulated environments, the approach incorporates data from actual prosthetic use by people navigating unpredictable real-world conditions. This method potentially exposes training algorithms to a wider variety of objects, surfaces, angles and handling challenges than laboratory settings typically provide.
The collaboration reflects broader trends in robotics and artificial intelligence toward learning from diverse, realistic datasets rather than idealized conditions. As robots move beyond simple repetitive tasks into more complex, adaptive roles, training data quality and diversity become increasingly important factors in system performance. The convergence of assistive prosthetic technology and industrial automation may accelerate progress in both fields, creating devices that better serve human needs whether as personal aids or workplace tools.
