The General Swarm Analysis looked at the physical limitations on a swarm. The basic restriction is that the swarm needed to fit in a 1.5 m cube on the starting platform. The conclusion was that a seven robot swarm is physically feasible, i.e. about the smallest dimensional robots that would work.
This analysis starts with the assumption that there are seven robots with two of them searching for samples. The other five, collector robots, pick up the samples and return them to the starting platform. Note that the SRR rules do not require the robots return to to the platform per se. Only the samples need to be on the platform at the end of the competition time.
The number of collector robots is not arbitrary. There are ten samples that can be collected. With five collectors each is responsible for picking up 2 samples. This could be done two ways.
Each collector uses its picker to collect a sample, continues to hold the sample in the picker, returns it to the starting platform, and deposits the sample. It then goes to pick up its second sample.
Each collector uses its picker to collect a sample and places the sample in a storage bin. It then goes to another sample, holds it in its picker, and returns to the starting platform. The collector could either deposit both samples on the platform or mount the platform itself. Either is acceptable to the rules.
A challenge when depositing samples on the platform is that the samples are kept apart. This is the contamination rule.
The use of five collectors greatly simplifies collection since no complex storage mechanism is needed. With a smaller number of collectors, up to a single collector, a large volume is needed to store samples - up to ten for a singleton collector. That storage needs to be accessible to the judges so they can retrieve the samples at the end of the competition.
The number of searchers is more arbitrary based mainly on a feeling that having more than two would become overly complex. But it would be physically feasible to have a swarm of eight robots with three searchers.
Having more than one searcher is driven by the time limit of two hours for the competition. The Area Analysis shows that it is a real challenge to search the entire area in that time. Having a swarm, in general, is an improvement since the searcher is no longer delayed by going off the search pattern to a sample and by the time to pick up a sample.
There is an interesting question: How late in the period can a searcher find a sample and have it collected. Assume the searcher finds a the 10th sample at the far reaches of the competition area. How long will it take a collector to get to the sample, collect it, and return it to the starting platform? That is the latest time during the competition where a sample can be usefully detected.
The first issue is where is the nearest available collector robot? This depends somewhat on the strategy used in the design of the collectors but let's assume that a usable collector is within a minute (120 m at 2 m / sec). I'm assuming that usable collectors will shadow the searchers throughout the competition for they are nearby when a sample is located. I'll allow a minute to actually locate the sample and another three minutes for the pickup. That is five minutes to get the sample.
How far might the sample be from the starting platform, worst case? That is hard to predict given that the competition area shape is unknown. For a square, the diagonal, the maximum length that can be traveled, would be 400 m. That means 800 seconds of travel time for an additional 14 minutes (rounding up).
Let's say that the last sample must be found 20 minutes before the end of the competition. That leaves 100 minutes for the searchers to do their job.
Repeating the basic calculations from the Area Analysis page, the searcher has to examine (80,000 m2 / 6,000 secs / 2 searchers) 6.67 m2/sec (22 ft2/sec). That is an area 2 m x 3.3 m (6.6 ft x 11 ft). This means the sample must be seen in an area about the length, but only half the width, of a good size bedroom. This is a pretty reasonable task.