A leading robotics company, renowned for its impressive demonstrations of dexterous robotic hands, has recently introduced a novel model designed specifically for machine learning research, jointly developed with Google’s esteemed DeepMind division.
With a rich history spanning over two decades, London-based Shadow Robotics boasts an impressive track record in robotic design, having worked with prominent organizations such as NASA, ESA, OpenAI, Google, and esteemed institutions like MIT and numerous universities throughout its illustrious career.
While earlier iterations of the world’s most dexterous humanoid robotic hand were seemingly familiar to humans, the novel Shadow Hand, co-developed with Google DeepMind’s robotics team, boasts a distinct design featuring only three fingers arranged in a grippable configuration.
“A crucial challenge in AI and robotics lies in designing hardware capable of performing advanced tasks with dexterity, while also being robust enough to support machine learning.” “Artificial intelligence systems learn through trial and error by verifying hypotheses in real-world scenarios, often performing actions within the limits of their capabilities.” This could trigger harm to the critical hardware, incurring costly repairs that would jeopardize ongoing experiments and potentially delay valuable scientific progress.
The novel robotic hand is engineered to withstand a considerable amount of abuse, including aggressive force demands, abrasion, and impacts.
What’s the latest scoop on your new hand?
Measuring 350mm in size, this item boasts impressive dimensions of 165mm huge and 160mm excessive, equivalent to approximately 13.78 x 6.5 x 6.3 inches. A single human finger tips the scales at approximately 1.2 kilograms, or 2.6 pounds, while the full hand, comprising all five fingers, registers a collective weight of around 4.1 kilograms, equivalent to roughly 9 pounds. Does it require a 48-volt, 200-watt energy provider?
The robotic hand reportedly excels at precise torque control, with each finger capable of generating up to 10 N of gentle yet deliberate fingertip pressure. The four joints of each finger are driven by motors situated within the base, connected via “tendons” that enable the fingers to transition seamlessly from fully open to closed configurations in just 500 milliseconds.
Every finger functions as an autonomous module, housing a diverse array of three-degree-of-freedom tactile sensors on its proximal and central segments. Additionally, a stereo digital camera system is strategically positioned within the silicone-based skin covering the fingertip, providing real-time high-resolution and wide-dynamic-range tactile feedback, which collectively enable the robotic platform to intuitively perceive its surroundings through the sense of contact.
New Hands on Deck: Spring 2024 Preview
What’s Next for Our Team?
In the event that one of the numerous finger modules sustains fatal damage during intense AI-driven experiments, it can be detached from the base module (which interfaces with a robotic arm) and replaced with a new unit, minimizing downtime in the process. If desired, the tactile sensors could be removed or replaced, enabling the communication network within the finger to detect the presence or absence of a sensor and transmit relevant information to a central computer system seamlessly.
While we don’t currently disclose pricing information for the Shadow Hand, our team is excited to unveil this innovative product to the public for the first time at ICRA 2024, opening its doors next week in Yokohama, Japan.
Supply:
