Behind the Fiction: The science of Robopocalypse

Daniel Wilson's much-anticipated novel Robopocalypse hits bookstores this week. It's the hyper-realistic story of a robot uprising, and is already being made into a Steven Spielberg film slated for 2013 release. Author Wilson gives us a brief excerpt from the opening of the novel here, then explains the scientific research that went into creating a battle scene unlike anything you've ever experienced.

Excerpt from Robopocalypse

Twenty minutes after the war ends, I'm watching stumpers pour up out of a frozen hole in the ground like ants from hell and praying that I keep my natural legs for another day.

Each walnut-sized robot is lost in the mix as they climb over each other and the whole nightmare jumble of legs and antennae blend together into one seething murderous mass.

With numb fingers, I fumble my goggles down over my eyes and get ready to do some business with my little friend Rob, here.

It's an oddly quiet morning. Just the sigh of the wind through stark tree branches and the hoarse whisper of a hundred-thousand explosive mechanical hexapods searching for human victims. Up above, snow geese honk to each other as they glide over the frigid Alaskan landscape.

The war is over. It's time to see what we can find.

From where I'm standing ten yards away from the hole, the killer machines look almost beautiful in the dawn, like candy spilled out onto the permafrost.

I squint into the sunlight, my breath billowing out in pale puffs, and sling my battered old flamethrower off my shoulder. With one gloved thumb, I depress the ignite button.

Spark.

The thrower doesn't light.

Needs to warm up, so to speak. But they're getting closer. No sweat. I've done this dozens of times. The trick is to be calm and methodical, just like them. Rob must've rubbed off on me after the first couple years.

Spark.

Now I see the individual stumpers. A tangle of barbed legs attached to a bifurcated shell. I know from experience that each side of the shell contains a different fluid. The texture and heat of human skin initiates a trigger-state. The fluids combine. POP! Somebody wins a brand new stump.

Spark.

They don't know I'm here. But the scouts are spreading out in semi-random patterns based on Big Rob's study of foraging ants. The robots learned so much about us, about nature.

It won't be long now.

Spark.

I begin to back away slowly.

"C'mon, you bastard," I mutter.

Spark.

That was a mistake: to talk. The heat from my breath is like a beacon. The flood of horror surges my way, quiet and fast.

Spark.

A lead stumper climbs onto my boot. Gotta be careful now. Can't react. If it pops I'm minus a foot, best case.

I should never have come here alone.

Spark.

Now the flood is at my feet. I feel a tug on my frost-covered shin-guard as the leader climbs me like a mountain. Metal-filament antennae tap, tap, tap along, questing for the tell-tale heat of human flesh.

Spark.

Oh Christ. C'mon, c'mon, c'mon.

Spark.

There's going to be a temperature differential at my waist level, where the armor chinks. A torso-level trigger-state in body armor isn't a death sentence, but it doesn't look good for my balls, either.

Spark. Whoomph!

I'm lit.

Behind the Fiction: The science of Robopocalypse

Writing Realistic Robots

My goal for Robopocalypse is to depict a terrifying, hyper-realistic future in which our familiar technology has run amuck and then evolved. Toward that end, I found inspiration and guidance by drawing on real-world robotics research.

I'm not trying to sound like a pretentious ass hat by saying that a lot of real robotics research went into my writing. I agree that science fiction is full of great, realistic robots that came straight from the imagination. All I'm saying is that my own creative process was to build on existing research in order to provide grounding for the dozens of unique robots that spy, stalk, and fight through the Robopocalypse.

Here's a glimpse of that process. The following is a breakdown of the decisions that went into the creation of "stumpers," those nasty little bugs you just read about.

Overview

The "stumper" is a crawling landmine. It is described as an insect-like, walnut-sized hexapod with filament antenna and a bifurcated shell. Stumpers typically hide themselves in cold environments and emerge in swarms upon detecting warm, bipedally-walking targets. Upon reaching a skin-temperature trigger-state, the stumpers self-detonate to devastating effect (hence the name).

Robotic Platform

The stumper is designed to injure and terrify human beings over a wide range of hostile outdoor environments for very little cost. The intelligence of a stumper is built into its design, which is complex, and not its behavior, which is simple.

Hardware

Stumpers are mass-produced from cheap materials. The body of each stumper is fashioned out of a few pieces of stiff fabric, similar to the biomimetic hexapod "cockroaches" built by Bob Full's POLY-PEDAL Lab at U.C. Berkeley [1]. This creates a robust platform that can locomote over varied terrain, resist crushing forces from being air-dropped into action, and consume minimal energy (crucial for a device that may wait for years before springing into action).

Sensors

Detecting body heat is accomplished via modified passive infrared sensors (PIRs). Gait recognition is accomplished via filament antenna attached to microphones that have been gated for desired vibrational frequencies.

Detonation Mechanism

Upon detecting a skin-temperature trigger-state, a stumper self-detonates by releasing stored current from a capacitor, vaporizing a wire "plug" that separates two liquids, which then explode on contact. My inspiration came from observing the fail-safe mechanism used by autonomous underwater vehicles. Upon catastrophic flooding, a plug dissolves in the salt water, releasing emergency ballast and surfacing the vehicle [2].

Behavior

Stumpers operate in structurally homogenous swarms with identical behavioral strategies. Self-organization arises as each stumper obeys simple local rules, similar to fish schooling or birds flocking [3]. For example, each robot stays a certain distance from the others while target seeking. The result is an emergent behavior in which stumpers flow around obstacles and surround prey. This decentralized approach is reactive and requires no active communication; it minimizes processing by mapping sensory information directly to action; and it requires no leader or central control [4]. The loss of an individual stumper causes no disruption to the behavior of the swarm.

Placement

Limited numbers of stumpers must distribute themselves to maximize the probability of human contact. Distribution of a living minefield requires only local communication, as shown by researchers on the "Self-healing Minefield" project [5]. Through passive communication, each stumper keeps a set distance from its neighbors, resulting in tunable placement over a given area.

Target-seeking

The systematic exploration of an area for resources by non-communicating agents falls under the umbrella of a "foraging problem." Ant species have evolved numerous effective solutions which have been replicated in autonomous multirobot systems. Typically, these solutions result in randomized fractal paths spreading outward from a central focal point [6]. In addition, using body heat to find human targets is a common "search and rescue" problem for robots designed to operate in disaster zones [7], including snake robots designed by the Biorobotics Laboratory at CMU [8].

Communication

Passive communication between the robots is accomplished using vibration detection. While in motion, stumpers generate a vibration tailored for easy recognition by nearby comrades. This form of vibrational communication is modeled on mating crickets, who emit frequencies that are strong enough to maximize range, yet stealthy enough to minimize the threat of predation [9].

Summary

And that's a brief overview of the sort of decisions that went into creating realistic robots for Robopocalypse. Luckily for the reader, very few of these details explicitly make it into the book. However, this level of realism is in my head as I write. Just proof that no matter how boring they seem, technical papers and research presentations can do wonders for the imagination.

Citations

[1] Dudek, D. M. and R. J. Full. 2006. "Passive Mechanical Properties of Legs from Running Insects," Journal of Experimental Biology, v. 209, pp. 1502-1515.

[2] Proenza, J., Ortiz, A., Bernat, G., Oliver, G. 1999. "A Cost-effective Hardware Architecture for Fail-safe Autonomous Underwater Vehicles," The 6th IEEE International Conference on Electronics, Circuits and Systems, Pafos, Cyprus, pp. 1487-1490.

[3] Hackwood, S. and Beni, S. 1992. "Self-organization of Sensors for Swarm Intelligence," 1992 IEEE International Conference on Robotics and Automation, Nice, France, pp. 819-829.

[4] Balch, T. and Arkin, R.C. 1998. "Behavior-Based Formation Control for Multiagent Robot Teams," IEEE Transactions on Robotics and Automation.

[5] Rolader, G.E., Rogers, J. and Batteh, J. 2004. "Self-healing minefield", Proceedings of SPIE, v. 5441, pp. 13, doi:10.1117/12.547923

[6] Goss, S., Beckers, R., Deneubourg, J., Aron, S., and Pasteels, J. 1990. "How Trail Laying and Trail Following Can Solve Foraging Problems," Behavioral Mechanisms of Food Selection, ed. R. Hughes, Springer-Verlag, Hidelberg, Germany, pp. 661-78.

[7] Markov, S. and Birk, A. 2007. "Detecting Humans in 2D Thermal Images by Generating 3D Models," Advances in Artificial Intelligence, ed. Hertzberg, J., Beetz, M. and Englert, R. Springer-Berlin, v. 4667, pp. 293-307, doi:10.1007/978-3-540-74565-5_23
[8] Wolf, A., Choset, H. M., Brown, H. B., and Casciola, R. 2005. "Design and Control of a Mobile Hyper-Redundant Urban Search and Rescue Robot," International Journal of Advanced Robotics, v. 19 (8), pp.221-248.

[9] P. Hill. 2008. Vibrational Communication in Animals. Harvard University Press.