I don't have the tools available to confirm a negative result, but it would be good to see this settled one way or the other (if it hasn't already; has it?).
Suppose a single glider hits a constellation of non-moving objects, (call it A) that lies in a kXk bounding box, for some small k<=20 for instance. The collision emits at least one additional glider and settles into a new constellation B. A single collision with B also produces another glider and settles into constellation C. Eventually, such a series of collisions restores constellation A (at the same place or maybe not; if rotated or flipped, it could be brought there by more collisions; or it might be shifted). Note that it is possible that A=B. There is no upper or lower bound on cycle length; there just has to be a cycle. B, C, etc. do not have to fit in the box. Just one step in the cycle does.
The reason for restricting k is that eventually a reflector meets the definition (snark is 23x17 and even p30 buckaroo does not violate the above definition, which does not mention period at all). Extending the bounding box much farther, you get into universal constructor territory. For this question, k is small, maybe k=15.
Another point to emphasize is that each collision must emit at least one glider. Otherwise, there are many glider-consuming solutions to restoring objects and constellations.
After revisiting some old search ideas and colliding gliders with natural junk, I'm convinced that there are only two constellations or objects that satisfy the above (up to equivalents like adding inert parts to the constellation).
One is the remarkable biloaf collision. The other is this glider shifter I found using primitive and forgotten methods in 1994.
Code: Select all
x = 81, y = 80
2bo$obo$b2o67$72b2o5b2o$72b2o5b2o5$79b2o$79b2o$67b2o$66bobo$68bo!
I thought maybe there was something else out there consisting of a natural constellation in a 15x15 box that could be reset through several steps like the above. I don't think so anymore based on this search (quoting myself from another thread).
(1) I began with millions of random starting seeds, generated them to find a stable ash with 20x20 bounding box constraints, normalized their orientation and phase (with blinkers), removed duplicates and obtained a list of 49905 "natural" stable and period-2 targets.
(2) Set up all glider collisions giving the glider 30 steps before hitting, which is usually enough to make it work coming from infinity.
(3) Tested these collisions for output gliders and non-zero stable ash fitting in a 20x20 bounding box.
(4) Collected the ash and dedupped again to use as another round of targets.
Repeat above until the set of targets is stable. There are some other things you can do to remove junk, such as taking the intersection of the target list from current and some past round.
The point is that ultimately a target is only reusable if it consists of collision ash. However, even these may be uninteresting as so many collisions produce ash that does not ultimately lead to a cycle.
So my conjecture is that the 3-block glider shifter and the biloaf are the only objects or constellations that can fit in a 15x15 box and be reset to the original shape without glider loss (or you can add the empty pattern to this list if you like). (EDIT but this is probably incorrect, if you let the intermediate constellations get large, assuming you can engineer a cleanup using only glider-producing collisions)
Is there a simple way to test this with modern tools? The basic idea is that there is a finite set S of stable constellations that fit in a 15x15 box. Non-lossy glider collisions transform this into another finite set of ash constellations (which need not fit in the box). Does this transformation graph contain any cycles other than noted above? My belief is no, but it would be great to see it verified and of course even better to see it refuted.
If it makes the problem more tractable, we could have some restriction on the constellation to rule out exotic, dense still lifes that are almost certain not to be part of a cycle. Anything that would work should come up in soup results. It is just possible that some Bellman-derived still life could fit in the box, and work as a catalyst. Rather than try to make the negative result apply, I suggest placing reasonable restrictions on the allowable constellation.