For our 3+1-dimensional space time, the Lorentz-group is six-dimensional: three dimensions for rotations + three dimensions for velocities = 6 dimensions. I found that there are extremely simple local rules based on which a very large number of simple 4-dimensional gird-graphs can be interlaced with each other in order to form a graph, such that the geodesic-distance of this graph perfectly approximates the Lorentz distance in 4D, resulting in the emergence of isotropy from discrete Wigner-rotations. Each of the grids corresponds to a point in the Lorentz-group. These points are arranged such that they completely fill out the six-dimensional Lorentz-group homogeneously. For an example of such a simple construction see my archived post: https://web.archive.org/web/20200922042112/https://community.wolfram.com/groups/-/m/t/1953906 This concept should find its way into the Wolfram physics project at some point.

Given the transportation rule for distance, no gridlike representations of empty space would work. Instead a graph should be "sufficiently irregular" to produce isotropy on large scale.

Another guess: a wave equation modelled on a square grid looks more or less isotropic. Its finite-difference scheme can be seen as a sort of local update rule (and maybe combined with branching). However it implies storing values on graph nodes, and possibly has relativity issues.

Oh, so that's probably equivalent to Jonathan's model

Maybe the isotropic hard limitation of the speed of light can be saved, if resistances associated with spacelike grid edges are positive, and resistances associated with timelike edges are negative.

Neat idea, but I'm not sure that I like the concept of having hypergraphs with such large edge density! (not least because it leads to an unbounded Hausdorff dimension for space...)

See my comment on this related post (https://community.wolfram.com/groups/-/m/t/1950834), or read my Q&A answer (https://www.wolframphysics.org/questions/spacetime-relativity/why-do-you-get-a-euclideanriemannian-metric-as-opposed-to-a-taxicab-metric-induced-on-your-hypergraphs/) for an explanation of how we actually derive Riemannian (and, indeed, Lorentzian) metrics in the context of our models - it is, unsurprisingly, not just the combinatorial distance metric. Or have a look through my relativity paper (http://wolframphysics.org/technical-documents/) for some more complete mathematical details :)

See here for the 4D space time version: https://community.wolfram.com/groups/-/m/t/1953906

Yes it is integer valued, think of very large integers. If you use this distance to for instance measure the 4 sides and 2 diagonals of a very large Quadrilateral, then the ratios between these 6 lengths will approximate what you'd expect from Euclidean geometry, if you let the size of the Quadrilateral as well as the number of grids goes towards infinity.

The Spacetime is a directed acyclic graph (DAG), so there exists a longest path along the arrows. Paradoxically the Lorentz distance is the longest path between two nodes. This longest path is the one where the largest amount of time passed between two events, which is also a geodesic of this flat spacetime.

Blockquote Are you sure that the underlying topology is still a 2d grid. If you tie infinitely many grids together in that way. You loose the property of dimensionality as Wolfram defined it since a single step adds infinitly many neighbors making the dimension undefined.

Depending on how exactly you connect the grids, the number of direct neighbors per node can go towards infinity but it can also stay a small number despite the number of grids going towards infinity. But anyways, as far as I remember the way Wolfram defined dimensionality is based on how the volume of a ball grows w.r.t. its radius, where ball is a set of nodes that are at most at the graph distance of the radius and the volume is the number of nodes in the ball. But this is just one kind of definition. You could instead take a set of random nodes and measure all the distances between them. Based on this set of long range distances you can see how many dimensions there are. It's like a complicated trigonometry exercise. A simple example: take four points on the graph, measure the 6 distances. From this information construct a tetrahedron in a real 3D space. If the tetrahedron looks flat, then the graph probably represents a 2D space.