What do you get when you combine billion dollar hype, demo videos, and the ability to build robots which can’t even dust the shelf without neonatal care from humanbeings? A dump of regrets, and at least they will be in the shape of humanoid, is the growing concern in the Humanoids Summit, which took place at Mountain View in California.
However, despite claims of a robot army from Elon Musk and his pre-orders for his $20,000 “robot butlers,” which will be trained by their owner, there are some voices of concern in the tech sector. There is a lot of good tech work being done, a lot of smart people working on these, but they are not yet well-defined products, said Kaan Dogrusoz, CEO of Weave Robotics, talking about full-body humanoid robots. They are still more like the Apple Newton, a technological wonder of its times, and failed miserably due to their untimely launching.
But the truth is that it is quite expensive to create safe humanoid robots, to say nothing of their competence levels. Consultant Ani Kelkar of McKinsey placed a price on safety:In a typical environment, an average of $80 needs to be spent solely for the purpose of avoiding harming anyone in the process of effectively using the robot within a cost of $100. It is purely an effect of design complexity with regards to safety in human robot interaction. It becomes much more unpredictable due to Humanity.
Additionally, Pras Velagapudi, the CTO at Agility Robotics, We’ve been trying to figure out, not only can we make a humanoid robot, but can we make a humanoid robot that does useful work, he explains.For example, the Digit from robotics agility has been enhanced to stand 5’9”, accompanied by the capacity to lift 35-pound boxes. Equally important, it incorporates AI balance control and object recognition. The Digit is able to navigate human-designed spaces without any alterations, contrasting other robots that employ the use of wheels, but it remains in industries where regulations are well understood. Another field of research that aims to fill the gap between concept demonstrations and application circles is teleoperated assistive autonomy.
Mixed reality interfaces have been developed by various teams of researchers that allow human operators to control humanoids such as Nadia with the aid of VR headsets, and Probabilistic Movement Primitives (ProMPs) and Affordance Templates (ATs) enhance the manipulative tasks done by the humanoid in real-time processing. The shared control approach makes possible accurate action, human error correction, and smooth transitions between autonomous and human-controlled tasks. They achieved a positional error of 2.35 cm RMS while reproducing human-controlled tasks, which is crucial during tasks like opening doors or bimanual manipulation tasks.
However, teleoperation as a practice in commercial humanoids has remained controversial in nature. This is due to the consideration that has been raised in the discussions within the sector in relation to product demonstration and its impact on investor and consumer perceptions with regard to system autonomy as a result of not incorporating human-in-the-loop control in the system. The use of VR tele-op systems is vital within the sectors of training and backup in unstructured environments.
There is positive advancement in locomotion and manipulation technology. Today, actuators are more energy-efficient, compact, and cost-effective, and advancements in battery technology have led to increased lifetimes, and AI-driven control algorithms enable robots to dynamically weigh and grasp objects in real time. PAL Robotics and Invento Robotics are reaching out to society for implementing humanoids in their application domains and healthcare, while Optimus from TeslaLabs and Digit Robots from Agility provide scalability for commercial deployment. However, from the success stories of Jibo and Cozmo, social robots, it has been learned that merely solving complex technological hurdles for robots will not be sufficient for commercial success in the marketplace.
The take-back for the engineering community from this event was that humanoid robots must make amends for themselves before making their entry into the commercial market on a broad, mass scale. What this will include is appropriate safety engineering, appropriate disclosure of their capabilities, and selective applications where their humanoid form gives them an unequivocal edge. It will otherwise, like Newton’s, meet its end in haste in a manner that has absolutely nothing to do with innovation but, altogether, has to do with misconstruing what it takes for a product to be complete, for what otherwise remains, like Newton’s, but a prototype.
