Swarm or Singleton?
The competition rules allow for the use of a swarm, e.g. multiple robots, instead of a singleton robot. As we saw in the Area Analysis, a singleton robot is hard pressed to search the roving area in the allotted time. How might a swarm, or a swarm and singleton cooperating, operate and what are its challenges?
One assumption made in any of the swarm possibilities is that coordination occurs among the robots. The robots need to coordinate on the area searched by the individual units, where they may be searching next, and where samples were found and if they were successfully returned to the base. After a general analysis I'll analyze the coordination issues.
Here are the possibilities I see:
- Singleton - a single robot that searches for samples, collects them, and return them to the base,
- Swarm - multiple robots that each search for samples, collect them, and return them to the base,
- Singleton search / Swarm collect - a single robot does all the searching while the swarm collects and returns samples to the base,
- Swarm search / Singleton collect - the swarm searches and while the singleton collects and returns samples to the base.
- A hybrid of of the last two with multiple search and multiple retrieval robots.
The biggest drawback to the singleton is the time constraint. Doing everything with one robot in the time allowed is just a challenge. A minor drawback is that a singleton is mechanically more complex making it more difficult to design and implement. One specific point of complexity is the need to store samples in separate bins to maintain sterility. Note, I am not considering the option of picking up a sample, taking to the base, and resuming the search because of the time limitations. (Actually, I should analyze this more. It might be more valid than it first appears.)
For the swarm only option, each robot is a smaller version of the singleton. Each robot searches and when it locates a sample returns it to the base. It then resumes the search. This option takes care of the time limitation by doing the search in parallel. It reduce mechanical complexity through the elimination of the storage bins. But basically, each of these robots is a singleton built at a smaller scale which probably increases the mechanical complexity. A big assumption is that, say, a five robot swarm would each find two samples and return them to the base. Each robot would have to find and return a sample in an average of 720 seconds (12 minutes).
The next to options simplify, in different ways, the mechanical complexity while still providing parallelism in activities to address the time limitation. In both these swarms the search and retrieval functions are separated.This means an individual will only have cameras for searching and another robot only has the retrieval mechanisms. Each robot is more specialized which means each individual is less complex.
With the singleton search and swarm retrieval, my initial reaction is that putting the more complex retrieval mechanism on multiple robots is not a good approach. On second thought, perhaps the retrieval mechanism is greatly simplified since each robot only has to handle a single sample at at time. Or maybe two samples?
The singleton retrieval and swarm search requires the complex retrieval mechanism with the sterile bin capability on the singleton. The search robots are smaller and each can look more closely at the area it is searching, i.e. a 'wing' length something less than 3 meters (see Area Analysis).
As I am writing this in September 2012 the swarm concept is becoming quite intriguing. All my previous thinking was based on a singleton handling all the tasks. That greatly influences my area analysis. Fortunately, the information developed there should be readily applicable to swarm analysis. As a result, I'll end this page to go into analyses of aspects of swarms.
The General Swarm Analysis looks into the size, weight and other physical limits on swarms.
The Swarm Activity Analysis looks into how the swarm would search for samples and return them to the starting platform.
The Swarm Communications Analysis looks at the requirements for communicating among the swarm and how the Home Beacon might be used to facilitate swarm activity.