I did say a couple of days ago I would post an idea for a moon lava tube/pit probe with a sketch or two. For now we could call it a Spike Sphere Probe, the fact is there would be three types of Spike Sphere Probes working together as a team:

Sphere model 1 = probe Sphere model 2 = telemetry relay Sphere model 3 = power sphere

First I will explain how the probe moves. The Spikes are telescopic, maybe two piece and overall length of around twenty cm. They would retract and extend some ten cm in each direction, giving a total of some twenty cm between full extend and full retract. The Sphere itself would be around one meter or more in diameter. To make the sphere move a determined number of Spikes would extend with intention to push the sphere, those Spikes that proceed the extended Spikes would retract. This action would repite itself in a ripple fashion towards the desired direction. In the tips of some of the Spikes would be certain sensors such as thermal resistance, floor/wall temperature, radiation measurements and so on, infrared light emiters and recievers could guide the sphere. It,s battery would be wireless recargable. The sphere would be constantly transfering data to the relay un Its.

Each relay sphere would not only transfer to the next relay but would also recieve the data necessary to follow the probe.

The power sphere is exactly that, it too would follow the probe, when the probe lacks energy it would recieve energy from the power sphere.

Later I will continue to explain

Interesting concept.

I don't know if "spikes" are best suited for the moon, something with a larger footprint might be more suited to the dust - keeping this out of the telescopic components might prove to be an issue long term.

The propulsion method generally seems viable - that's a lot of actuators to move all the spikes tho... to think of a 1meter ball, and spikes, I'm thinking 30-50 spikes to get good movement in any given direction. Unless you have some cunning method to push them out from the inside by a gyroscoped mechanism. That could take a lot less actuators.

To assume there's some vague intent to have these mass deployed for surface(and by topic of your post, assumed subsurface) exploration then to relay telemetrics from every unit would rapidly consume massive amounts of bandwidth. Instead, log and recycle the logs and have provision to obtain on demand. Unless you don't expect it to come back. Thusly operators or other spike sphere can use some sort of API to request the specific data they require, when they require it, from where they require it. Less total transmissions.

What powers the power sphere?

IR reflection might be good as a distance sensor, in the tips of spikes, but unless this sphere is clear I don't see many ways to get that reflection into a CMOS - a camera at the tip of spikes will give multiple redundant copies, but massively increase cost and weight over having a single sensor, or small cluster. It'd also place them in the most likely (IMHO) place to get damaged. The other method that sprang to mind would be multiple camera on the outside of the sphere - and hope they don't pack up with moon dust. This would also give mulitple redundant copies, and if you got the sphere rolling fast enough, and switched camera feed fast enough, you might be able to get a constant "straight ahead" (or any other angle) veiw. With some cunning software you could compile footage that operators could watch without strain or discomfort.

I assume these things would be behaving somewhat like worker ants, plotting up a nest site?