Putting Robots to the Test

July 2012
By Bryan Bergeron

If you’re a fan of Blade Runner, Battle Star Galactica, Alien, Terminator, or Prometheus, then you know that — at least for most people — the pinnacle of robotics is much more than a walking and talking tin can. The robots featured in these and other Sci-Fi classics certainly pass the Touring Test: a necessary but inadequate measure of how closely a robot resembles a human.

Although not explicitly stated, these robots are physiologically correct – they breathe, bleed, and sweat as we do. Short of surgical exploration, the ultimate android is physically indistinguishable from a real human. This suggests the physical equivalent of the Touring Test for androids. I suggest the following test: Allow a clinician to physically examine humanoid robots with the non-invasive instruments used for a traditional physical exam: a reflex hammer, blood pressure cuff, stethoscope, ophthalmoscope, and their hands – for 10 minutes. The clinician — without engaging in a conversation with the robot — must determine whether their patient is human, robot, or something in between.

Because no human is perfect, this test suggests that the robot must exhibit some combination of normal and abnormal physical findings. Perhaps the robot’s blood pressure is a bit elevated, there’s a touch of asthmatic wheezing audible from the lungs, or there’s a slight heart murmur.

So, if this is perfection in a physical sense, then how do we get from where we are now to the future of robotics? Science fiction writers and self-anointed futurists have the advantage of not having to create a detailed project plan for the realization of their visions. A more practical assessment of the trajectory of human-like robotics is to look at the progress in the development of human surrogates to train clinicians in both civilian and military scenarios. Task trainers – system-specific physical simulators — are increasingly used to train medics, nurses, and physicians on how to save lives and treat real patients. There are commercial and academic task trainers for applications ranging from learning to suture wounds, applying a tourniquet to stop massive bleeding, and interpreting heart and lung sounds to measure blood pressure, and delivering a baby.

Depending on the fidelity of these trainers and the availability of competing products, prices range from under $100 to $200K. Even at the upper price range, there is ample room for improvement. So, where to start?

I’d begin by reviewing the classic sci-fi films to decide what’s a worthy goal. Superhuman strength or the ability to morph into another object or person probably shouldn’t be at the top of your list. Take a rational assessment of your resources. Do you have a team of PhDs working for you or perhaps only a drawer full of servos and microcontroller boards? Assuming the latter, I’d start with something simple, such as replicating a basic reflex arc — think knee reflex.

If your neuromuscular system is normal, then when one of your tendons is suddenly stretched by someone else – say, by a nurse striking it with a rubber reflex hammer – then the muscle attached to the tendon will respond by quickly contracting. There is some value in this reflex if you examine what happens when you land after jumping down from a step stool. As you land, the tendon is suddenly stretched, signaling the muscle to contract, cushioning your landing. This unconscious reflex — which doesn’t involve the higher centers of the brain — helps keep you upright and on your feet. (By the way, if you hit the tendon yourself or consciously focus on the reflex, you won’t see it – do you
know why?)

To replicate the basic tendon reflex, you don’t need much more than a microcontroller, servo, and sensor. The sensor could be a commercial strain gauge or a piece of carbon-impregnated foam (the kind used to control static electricity). A sudden change in resistance for either one can be used to trigger the servo into action. While you’re planning the next phase of your R&D toward a humanoid robot, a practical application is a self-righting reflex for a walker or crawler. Once you understand and apply the reflex arc, you can move to more complex physiology — from muscle strength and fatigue, to the complexities of the visual system. SV