Summary

Bugs are general purpose robotic platforms designed for experimentation with ideas of swarm intelligence. They are physically embodied autonomous agents interacting as an artificial life system. Each robot is programmed with simple rules such as "seek light" and "avoid obstacles". Through their interaction with each other and the environment these simple low level behaviors give rise to interesting higher level patterns of organization.

Emergence

Emergence can be defined most simply as the property which makes some systems greater than the sum of their parts. It is the appearance of higher level order stemming from, but irreducible to interactions occurring between low level components. It is a property of systems composed of multiple interacting parts, for example a flock of birds, where low level refers to the behavior of a single agent in the system (a bird) and higher level refers to the behavior of the system as a whole (the flock). Irreducibility means that the behavior of the system as a whole is not deducable from the behavior of a single agent. Even with a complete understanding of the rules a part follows (as in robotic agents or other artificial systems) the behavior of the system is unpredictable. Emergence occurs in self-organizing systems which solve problems by drawing on masses of simple elements rather than a single executive. To return to the birds, no single bird is the leader of the flock. They are structured as decentralized networks rather than hierarchies. The rules of interaction between components of these systems takes place on the basis of local information without reference to a global pattern. In other words a single bird positions itself based on the birds in its immediate vicinity. It does not have an explicit internal representation of the flock as a whole. Other examples of biological systems which exhibit emergence include ant colonies, schools of fish, termites, bees, and packs of wolves. Emergence is also a property of many human systems including neighborhoods, traffic jams, the stock market, and the internet. On a more personal level, you are an emergent system as a multicellular organism; every one of your cells acts independently, and the higher level order of all these cells interacting forms you as a human being. ^

Why Does Emergence Matter?

Emergence represents a fundamental shift in the way we study the world around us. For the past three hundred years a reductionist philosophy of research has sought to understand systems by breaking them down into their constituent parts. It has assumed that if we understand how all the individual parts work we will understand the system as a whole; that high level phenomena are fundamentally reducible to low level organization. Emergence flips this view on its head, countering that systems level phenomena are fundamentally irreducible and are dependent upon multiple random interactions among components. The behavior of the aggregate cannot be derived from the properties of individual components. This impacts research occuring in many disciplines from biology to economics to philosophy. ^

Alife

What is life? How can we define what it means to be alive, and how do we seperate the living from the non-living? Many definitions of life exist incorporating ideas of self-replication, metabolism, evolution, and reverse entropy. One undeniable property of life is that it is emergent; no single element in a cell is itself alive, yet the interaction of the non-living elements it contains give rise to what we know as a life. (Irun Cohen) Artificial life is the study of artificial systems that exhibit behavior characteristic of natural living systems. It is the quest to explain life in any of its possible manifestations, without restriction to the particular examples that have evolved on earth. (Christopher Langton) It is the study of life as it could be rather than as it is known to be. A young discipline, artificial life is a split from traditional artificial intelligence research which uses centralized top-down models based on extensive sets of rules. The alife approach to intelligence works from the bottom up focusing on intelligence as an emergent property of a a system of interacting components. For example an AI robot might have a giant database containing explicit rule sets and operations for handling various situations, such as "if p then q", "if q then r", etc. An alife system on the other hand might use a genetic algorithm to evolve a solution to a problem, or a group of simple interacting robots to achieve a behavior. ^

Collective Robotics

The use of emergent models for problem solving allows a bottom-up decentralized approach to replace the top-down method oriented around a central intelligent agent. Emergent models disperse a problem to multiple autonomous agents and allow intelligence to emerge from their collective behavior. While most artificial life research occurs inside the computer as simulation, swarm-based or collective robotics is artificial life embodied. It takes real-world physics as both a challenge and an indisposable ingredient of emergent systems. Collective robots generally do not engage in direct communication, instead they act autonomously on the basis of local information from their changing shared environment. Inspired by social insect societies collective robots are not only a useful tool for investigating the dynamics of emergent systems, they are also often less expensive, more robust and more flexible than traditional AI robots.

Collective robots often make use of stigmergy, or coordination of activity through modification of their shared environment. For example, ant trail following occurs because indiviual ants excrete pheremones as they walk between a food source and their nest. Since they are in turn attracted to this pheremone excretion it generates a system of feedback where ants are both attracted to the trail and adding to it, making other ants more likely to follow it. The idea of stigmergy originated with the study of termites and other social insects but has since been applied to many other systems including the world wide web, viewed by some as the first stigmergic medium for human beings. Stigmergy is essentially a highly effective mechanism for biological self-organization, and in turn a tool for collective robotics. There are four main ingredients of self-organization:

1. Positive feedback (amplification) ex. pheremones
2. Negative Feedback (counterbalance) ex. saturation, exhaustion, competition
3. Randomness, allows for discovery of new solutions
4. Multiple interactions among individuals

Using these ideas as a guide many swarm based robotic systems have been created, tackling tasks such as clustering and sorting, self-assembly and reconfiguration and collective transport. I used these rules as a framework in developing various programs for Bugs as well. For example, in my seek light-coordinate walking program the positive feedback is the individual robot's illumination and attraction to light. As more bots converge in one location that area in turn becomes more illuminated and attracts more robots to it. There are two negative counterbalances in the system. First is collision, as the robots congregate they collide and must back off. Second is exhaustion, each robot has an exhaustion function and after a period of coordinated movement it becomes tired and must rest. Randomness in the system comes from random walks the bots take when they are seeking each other, and multiple interactions arise from the situation of ten mobile robots in a relatively small area.