Coordinate thousands of robots like a small world.
Spatial Robotics builds small-world networks for multi-robot coordination — balancing local efficiency against global awareness so a swarm reaches agreement in a few hops, not a full mesh. Design topologies, inject failures, and benchmark consensus entirely in your browser.
- Watts–Strogatz topologies
- Python · C++ · ROS2
- No GPU required
live · watts_strogatz · n=28
- Avg path
- 2.53
- hops between robots
- Clustering
- 0.27
- local cohesion
- Propagation
- 11.1×
- reach / tick
A handful of long-range shortcuts (in blue) collapse the network diameter — the small-world effect, live.
The platform
From graph theory to robots in the field
One core powers the design studio, the benchmark suite, and the on-robot SDK — so the network you prototype is the network you deploy.
Small-world graph construction
Watts–Strogatz, Barabási–Albert, k-regular, and our proprietary Spatial Small-World — tuned to balance local clustering against global reach.
- β rewiring control
- 4 topology families
- Spatial constraints
Message-passing & consensus
Average, max, and weighted consensus plus Hungarian and auction allocation. Watch agreement propagate hop-by-hop across the swarm.
- 7 coordination policies
- Bandwidth-aware
- Per-tick message ledger
Discrete-time simulator
Deterministic, seeded, tick-by-tick execution from 10 to 10,000 robots. Reproduce any run exactly — no GPU, no cluster.
- Seeded & reproducible
- Up to 10k robots
- Browser-native
Failure-injection benchmarks
Stress topologies with packet loss, robot dropout, and link latency, then measure how gracefully coordination degrades.
- packet_loss
- robot_failure
- link_latency
SDK — Python, C++, ROS2
Take the exact graph you designed in Studio to the field. Python, C++ (pybind11), and a ROS2 bridge are being built on one shared core.
- Python SDK
- pybind11 bindings
- ROS2 bridge
Live metrics & convergence
Track average path length, clustering, disagreement, and message volume in real time — the numbers that prove a swarm is coordinating.
- Convergence curves
- Path-length tracking
- Streaming telemetry
10,000+
Robots per simulation
Discrete-time, deterministic
4
Topology families
incl. Spatial Small-World
7
Coordination policies
Consensus · allocation · flocking
96%
Fewer messages
vs. full mesh — sample simulation
Benchmark suite
Reproducible by design
Every topology is scored on a reproducible benchmark: how many ticks to reach consensus, and how many messages it costs. In our simulations, small-world graphs converge about as fast as a full mesh while sending a fraction of the traffic.
- Seeded runs — anyone can reproduce the exact result.
- Built to rank reference and community submissions side by side as the beta opens.
- Scored under failure injection, not just ideal conditions.
Convergence leaderboard
ticks · lower is better
Sample benchmark data (simulation)
- 1Small-World 100 (β=0.12)128watts_strogatz400 msg/tick
- 2Full Mesh 100160k_regular9,900 msg/tick
- 3Scale-Free 200186barabasi_albert720 msg/tick
- 4Star 100220k_regular198 msg/tick
- 5Spatial SW 1k412spatial_small_world4,000 msg/tick
Build your first swarm in minutes.
Drop nodes on a ring, tune the rewiring probability, pick a consensus policy, and hit run. No install, no GPU, no cluster — then export the same graph to your robots.