Robotics and Nature
An insectoid robot is a small robot that may have some insect-like features which can include various methods of locomotion (running, leaping, flying) and navigation. These robots often implement artificial intelligence based on insect models.
What many people don't realize about robotic design of this genre is that robot designers are mostly counting on the insects themselves for inspiration as to how to solve various design and function problems they encounter along the way.
Here are just a few examples of some of the issues that scientists encounter in developing working prototypes.
Mimesis and robotics
Let's start of with locomotion, something that many robots today seem to present as facile results of robotics and programming. But all is not that which meets the eye.
Insects have articulated joints and to replicate their motor functions, this is not so simple as with a stickman figure. For any robotic replicant to function just as a hexapod (the form of an insect), there must be sensor-motor control that insects the neurology of a real insect which can perceiveand react tovarious types of terrain. Without this detail, no insect model or any amount of programming will keep our hexapod upright, much less moving. As per sentient creatures, the legs and joints of these hexapods must be controlled individually and they need to act in concert according to information received from their limb positions and load sensors.
Are you still with me?
Add to this needing to take into account the gait and speed of the insect and you have a cocktail for a series of coinciding actions that need to happen in less than a nanosecond in order for the hexapod to respond all the information that is incoming and all this information must be kept at speed with the artificial nervous system that transmits to the mechanical body what must be done: speed, gait, tilt, body positionality, obstacles to avoid, weight of body on turns, the spring stiffness of its legs, the coordination of all six legs with three on each side moving in concert.
This is why what work we do in robotics must be undertaken with other specialists from various disciplines from physics to aerodynamics to neurobiologists and biologists. For instance, biologists look at our models and decide what is accurate, what is not. Their feedback forces us back to the design board where we must tweak certain facets of our design and where along with robotics engineers and neurobiologist, we can improve the next prototype in understanding the way insects function to project this function into our latest model. Biologists then gain a platform upon which they can test their theories of insect motor control and if our model still doesn't pan out, we work on biologists' observations of our robotic models to improve the mimetic (representative) force of our work.
Artificial Intelligence Gone Wild!
Back in 1986 at MIT, researchers were already toying with the idea that
What is important to know regarding this scale of robotics is the position of Moore's law within the field. You see before the last decade, it was commonplace to anticipate that further improvements to microprocessors that would close a performance gap where size, speed, and motor agility would improved due to the developments of microprocessors. Before the past decade, processor development was keeping pace with Moore’s law, a theorem that predicted a doubling of the number of transistors in a dense integrated circuit (IC) about every two years. However, in robotics and computer science, it is widely accepted that Moore’s law is in its waning phase and thusly, we can no longer count on the leaps in microchip size reduction. As a result, we need to explore alternative approaches to both the computing hardware and the AI of small, autonomous robots.
This is also where insectoid robotics comes in handy. Researchers in his study, GCHE de Croon et al argue that even for neuromorphic processors, one should not simply apply existing AI algorithms but instead we need to exploit insights from natural insect intelligence to get maximally efficient AI for robot autonomy. What this means in plain English is that AI is now working on the modelling from biologists and neurobiologists to fashion future errors in biomechanics and robotics design. These researchers argue for the use of neuromorphic processors that no longer simply apply existing AI algorithms but instead that exploit insights from natural insect intelligence to get maximally efficient AI for robot autonomy.
Final Thoughts
While there are many debates within the field regarding Moore's Law and the need for transformers in future robotics, there are now rising debates about Huang’s Law which posits that the performance of GPUs and AI accelerators is improving at a rate much faster than the traditional two-year doubling cycle described by Moore’s Law. Form versus function versus hardware versus AI. It's all a beautiful mess for those of us on the frontlines trying to bridge technology with nature where reduced size is pivotal for design.
No matter which side of this debate proves to be correct, one thing is sure: AI is moving faster today than hardware development and this cannot be overstated. And the knowledge AI is gleaning on its own from insect physiology and physiognomy is outpacing hundreds of laboratories packed with scientists. And the benefits to humankind are immense.
From the dragonfly robots which surveil mosquito populations in regions harshly affected by dengue, chikungunya virus, dirofilariasis, malaria, and Zika to MIT's recent advances with robotic insects that may bring us to a new era in
And what comes next might just surprise us all.
