When three robots are better than one: Distributed Multirobot Systems
The 1972 big budget movie Silent Running should be the ultimate Earth Day sci-fi movie: a realistic multirobot system of 'bots are entrusted to save the world’s remaining trees transplanted to space orbitals. Instead of inspiring environmentalists, it literally inspired Mystery Science Theater 3K. Yet Silent Running is a must-see for any robotics aficionado because the robotics is accurate. Jump to the video podcast here.
In a nearish future, the last forests have been relocated by unappreciative, environmentally reprehensible human race to space stations, in effect, terrariums in space. The terrariums are tended to by three little legged robots (called Huey, Dewey and Louie after Donald Duck’s nephews) and some human misfits. When the bean counters on Earth tally up the costs of having forests in space that no one cares about (and can’t visit), they order their destruction. Fighting ensues between the a*hole crew members who are happy to nuke the forests and return home and the crazy, gone-all-Sierra-Club crew member Lowell who wants to disobey and save the forests. For anyone too dense to miss the pro-environmentalism plot, Joan Baez, the iconic folk singer who is normally subtly evocative, sings the most overtly cloying “save the trees” song ever.
Yes, this is a movie worthy of Mystery Science Theater 3000. And, according to imdb.com, it actually inspired the show.
Despite the dose of dreadful earnestness, Silent Running is actually pretty accurate about robots. The director, Douglas Trumbull, hot off of his special effects work with Stanley Kubrick on 2001: A Space Odyssey, was all about technical realism. Maintaining a space station in the harsh vacuum and radiation of space requires robots, as NASA learned decades ago, so bingo, Trumbull put in robots. (Good man!)
What is particularly accurate about the robots is how Silent Running portrays the multi-robot coordination. Multiple robots offer many advantages, some are that several robots can tackle tasks that a single robot can’t do (e.g., move something heavy) and can cover more area faster or simultaneously protect an area. Ideally many robots would be cheaper (economy of scale).
Writers tend to assume a centralized control paradigm for multi-robot systems, where one computer or robot controls all the robots in a master-slave type of arrangement. This is particularly convenient for plots where the centralized computer is out to take over the world, like V.I.K.I. in I, Robot. Centralized control of robots has many problems in practice. A big one is that if the centralized controller is down for maintenance or destroyed (say by a young Anakin Skywalker in The Phantom Menace), then all the robots become useless (jokes on you, Trade Federation! You may be able to control the galaxy’s monetary supply but you made a rookie mistake in robotics!).
AI researchers in multi-robot systems tend to focus on the distributed control paradigm, where each robot has independent intelligence and responsibilities. Distributed controls raises different challenges. One fundamental challenge is determining how independent, and what is the minimum intelligence, can the robots be to get the job done. Are the robots aware of each other and communicate directly with each other, like wolves yipping to coordinate a hunt? Or are they only indirectly aware of each other and communicate through direct perception of the state of the world or stigmergy? Stigmergy is a biological mechanism where the action of one animal leaves a trace behind that influences the next action taken by another animal, like ants leaving a pheromone trail behind for other ants to follow.
A really nice example of exploiting direct perception and avoiding complex reasoning is work done by Lynne Parker in box pushing. She had a team of robots push boxes that were too heavy for a single robot to push. Each robot would start pushing and could sense whether the box surface was flat to the touch or jutting in or out. If the box was jutting in, the little robot would push harder even though it didn’t know anything about how many other robots there were or how they were pushing - just stimulus, response. If the box was jutting away, the robot would slack off, thus giving all the others time to catch up. If the box was flat, then the robot would maintain its force and trajectory. Notice that the robots did not have to know how many other robots were involved or communicate directly with the others. Plus it would still work if a person was moving one side of the box- the robot doesn’t know or care.
Another question that a designer of a multi-robot team has to consider is whether the robots should be the same or different? Identical robots, or homogeneous robots, working together are often programmed to behave as swarms of insects. A swarm in robotics implies that individual robots emulate insect intelligence and have a simple mechanical design so that having hundreds or thousands of robots is economically feasible. Huey, Dewey and Louie were homogeneous but not a swarm as each was independently highly intelligent and directly communicated and negotiated with each other as to who did what when.
Yet another research challenge is determining in advance how many robots are needed to guarantee accomplishing a specific task. If there are too many robots working on a task, especially if they are acting as an insect-like swarm, it can actually degrade the overall task performance - an example of the old adage that “too many cooks spoil the broth.” In robotics, that phenomenon is called interference.
Of course, centralized control has a major advantage: it may be optimal on a global scale. In Silent Running, Louie, limited by its local perception of the situation, can’t perceive the bigger picture and ignores the advice from Lowell, who has a better, global view of the situation. Instead, Louie, like an animal or a person, trusts its own judgment. High realistic from a robotics programming standpoint and, sadly, highly realistic from a consequences perspective.
So how did Trumbull make Huey, Dewey and Louie? They seem so natural… Ah, here’s where maybe the movie is more for appropriate for discussion with older kids. The robots are multiple amputees walking on their hands. Really. It does afford an opportunity to flip the image of the disabled playing robots and discuss how robots could help the disabled… Check out the article about the "almost robots" on cyberneticzoo.com
Notice that when the robots stop to get directions from Lowell or other crew members, they sit down - in a robot, it would be natural to do this to conserve energy, in a person, it was for the actors to reduce fatigue. Likewise, the robots turn slightly to pay attention to the humans walking by or they turn and move in ways consistent with conversational poses and gestures.
Is Silent Running deserving of cult movie status? Maybe not for its overall contributions to apparatus theory in film studies, but definitely for its cinema verite’ incorporation of robots.
Huey and Louie - rest in peace, and Dewey keep up the good work in that Big Forest in the Sky. And someone, please, mute Joan Baez.
- Robin
If you want to learn more about distributed robotics, check out Chapter 17 in my textbook Introduction to AI Robotics, second edition.
If you are looking to learn more about multirobot systems by reading science fiction:
The Metal Monster appears to be the first book with multirobot systems
Prey of Gods is a funny take on robots and teenagers. And South African gods!
The most accurate scifi on swarms IMHO is Kill Decision
If you are hoping for a better movie,
Passengers is more watchable than Silent Running, but really a rom-com
look at the lists here