Start a Conversation
Active research frontierActive

Distributed Systems in Motion

Robots are one environment we are using today. The larger question is what happens when distributed participants move, disconnect, change hosts, act locally and still have to remain part of one coherent system.

The robot is not the product. The distributed system is.

Why this environment?

Not because we are becoming a robotics company.

Robots simply make distributed-systems problems harder to hide.

  • MobilityA participant is not where it was. Nothing about it may depend on where it was.
  • Intermittent connectivityAbsence is normal, and indistinguishable from failure while it lasts.
  • Heterogeneous participantsUnlike hosts, unlike runtimes, unlike capabilities, one system.
  • Partial knowledgeNo participant sees the whole. Every decision is taken from inside a fragment.
  • Local autonomyActing without asking, and still being answerable for what was done.
  • Changing topologyWho can reach whom is a fact about right now, not about the system.
  • Continuity under failureThe system outlives the processes, hosts and links that carried it.

Today the participants may be robots. Tomorrow they may be vehicles, machines, edge devices, mobile applications or systems we have not encountered yet.

Where we are exploring

Physical intelligence

Very crowded

perceptionmanipulationnavigation

Distributed system organization

Puppeteer exploration area
  • identity
  • authority
  • continuity
  • causality
  • commitments
  • consequences
  • heterogeneous participants
  • changing topology
  • durable history

Infrastructure

Crowded

ROSDDSnetworkingfleet managers

Puppeteer does not replace either layer. It explores what becomes possible between them.

We are deliberately working in the middle.

What Puppeteer brings

The existing research gives us the properties. The frontier tests what happens when the conditions get harder.

  • Identity

    Who is this participant?

  • Authority

    What may it decide?

  • Continuity

    What survives process, network or host changes?

  • Causality

    What keeps a consequence attributable to the act that caused it?

  • Commitments

    What obligations remain when participants disappear?

  • Consequences

    What follows from an action elsewhere?

  • Durable history

    What survives strongly enough to reconstruct the participant?

Now

Current frontier

Current question

What has to be true of a participant for it to remain part of one coherent system while it moves, acts on its own and loses contact?

Current test environment

ROS2 robots

We are currently using ROS2 robots because they combine mobility, local computation, sensors, independent action and changing connectivity in a compact experimental environment. The environment is the stress, not the subject: nothing on this page is a claim about ROS2, and nothing about the questions depends on it.

Current build

Nothing recorded yet.

What we are observing

Nothing recorded yet.

What remains open
  • Whether identity is a property of the participant or of the record the system keeps of it.
  • What a commitment means during the window in which no one can observe whether it was met.
  • Which decisions can be taken locally without the system losing the ability to say what happened.

Experiments in progress

One question each. An experiment that turns out to be wrong has done its job.

  • Exploring

    A Puppet Gets a Body

    Does the Puppet remain the same participant when its physical interface changes?

  • Exploring

    One Commitment, Two Hosts

    Can a commitment outlive the host that first carried it?

  • Exploring

    Authority Without a Permanent Coordinator

    Can participants coordinate authority without turning infrastructure into the organization?

  • Exploring

    Testimony Across Participants

    What changes when one participant acts on another participant's observation?

  • Exploring

    Continuity Through Disconnection

    Can organizational continuity survive topology changes and temporary disappearance?

Field Notes

Notes from active work will appear here.

Findings

Observations that survived their evidence, including the ones that went against what we expected.

No published findings yet. Experiments are still running.

From exploration to evidence

Not a pipeline. Most notes never become findings, and most findings never become papers — which is what makes the ones that do worth reading.

  1. Field noteSomething we saw, written down before we know what it means.
  2. ExperimentA question isolated enough to be wrong about.
  3. FindingAn observation that survived its evidence.
  4. DemoEvidence that stands on its own, executable.
  5. PaperAn argument the rest of the work rests on.

Latest activity

Activity will be logged here as the work moves.

The environment may change. The distributed-systems questions remain.

Identity. Authority. Continuity. Causality. Commitments. Consequences. Coordination without requiring a permanent center.

Research — the argument this rests onDemos — the evidence that already stands