Skip to content
RFL_GLOBAL
中文
  • SLAM-COKO // W7
  • Python
  • ANIMA

SLAM-COKO

A Robot Swarm That Shares One Map, Even After Comms Drop

Autonomous robot teams — mine-clearing swarms, building-clearance platforms, underground search-and-rescue units — must maintain a shared map of unexplored terrain. When radio comms are intermittent, each agent builds a partial map. Merging those submaps after reconnection without drift, duplication, or inconsistency is an unsolved engineering problem for most deployed systems. SLAM-COKO (codename: Amaterasu) implements CoKo-SLAM (arXiv:2604.00804), a compact keyframe-optimized approach to multi-agent Gaussian Splatting SLAM. Each agent maintains a photorealistic 3DGS submap.

MODULE STATUS: PRODUCTION

Multi-Agent Fusion

PROD

DIVISION
GENERIC
WAVE
W7
DOMAIN
GENERAL
WAVE 7 // ANIMA SUITE
FOUNDATION — SLAM-COKO
SLAM-COKO // W7 // 007/012
01THE CHALLENGE

THE PROBLEM WE SOLVE

Autonomous robot teams — mine-clearing swarms, building-clearance platforms, underground search-and-rescue units — must maintain a shared map of unexplored terrain. When radio comms are intermittent, each agent builds a partial map. Merging those submaps after reconnection without drift, duplication, or inconsistency is an unsolved engineering problem for most deployed systems.

Swarm-deployed units operating under electronic warfare conditions — where comms are unreliable by design — can re-merge their spatial knowledge the moment a relay link is re-established, providing a coherent photorealistic operational map without human reconciliation.

02THE SOLUTION

WHAT SLAM-COKO DELIVERS

SLAM-COKO (codename: Amaterasu) implements CoKo-SLAM (arXiv:2604.00804), a compact keyframe-optimized approach to multi-agent Gaussian Splatting SLAM. Each agent maintains a photorealistic 3DGS submap. When agents reconnect, loop closure detection identifies overlapping regions, and pose graph optimization fuses submaps with drift correction across the entire team.

CAPABILITIES

  • SLAM-COKO (codename: Amaterasu) implements CoKo-SLAM (arXiv:2604.00804), a compact keyframe-optimized approach to multi-agent Gaussian Splatting SLAM
  • Each agent maintains a photorealistic 3DGS submap
  • When agents reconnect, loop closure detection identifies overlapping regions, and pose graph optimization fuses submaps with drift correction across the entire team
  • The compact keyframe strategy keeps memory footprint bounded regardless of map extent.
03ENGINEERING

WHY THIS IS HARD

Building SLAM-COKO requires solving multiple coupled problems:

  1. 01Autonomous robot teams — mine-clearing swarms, building-clearance platforms, underground search-and-rescue units — must maintain a shared map of unexplored terrain
  2. 02When radio comms are intermittent, each agent builds a partial map
  3. 03SLAM-COKO (codename: Amaterasu) implements CoKo-SLAM (arXiv:2604.00804), a compact keyframe-optimized approach to multi-agent Gaussian Splatting SLAM
  4. 04Each agent maintains a photorealistic 3DGS submap

SLAM-COKO solves these through careful architecture design and rigorous validation.

04BENCHMARKS

PROOF, NOT PROMISES

Key performance metrics:

PROOF, NOT PROMISES
METRICVALUE
Multi-Agent FusionLoop closure + pose graph optimization across N agents — bounded drift on merge
Map FormatCompact keyframe representation — memory scales with scene complexity, not trajectory length
RepresentationPhotorealistic 3D Gaussian Splatting — renderable for human operator review
Defense AngleSwarm mapping under EW/comms-denied conditions, building clearance, underground operations
05BUILD STATUS

WHAT'S BUILT TODAY

2/4 COMPONENTS COMPLETE
WHAT'S BUILT TODAY
COMPONENTSTATUSNOTES
Multi-Agent FusionCOMPLETELoop closure + pose graph optimization across N agents — bounded drift on merge
Map FormatCOMPLETECompact keyframe representation — memory scales with scene complexity, not trajectory length
RepresentationIN PROGRESSPhotorealistic 3D Gaussian Splatting — renderable for human operator review
Defense AngleIN PROGRESSSwarm mapping under EW/comms-denied conditions, building clearance, underground operations
06APPLICATIONS

WHERE SLAM-COKO DEPLOYS

  • APP_01

    AUTONOMOUS SYSTEMS

    Swarm-deployed units operating under electronic warfare conditions — where comms are unreliable by design — can re-merge their spatial knowledge the moment a relay link is re-established, providing a coherent photorealistic operational map without human reconciliation.

  • APP_02

    RESEARCH LABS

    SLAM-COKO (codename: Amaterasu) implements CoKo-SLAM (arXiv:2604.00804), a compact keyframe-optimized approach to multi-agent Gaussian Splatting SLAM.

  • APP_03

    EDGE COMPUTING

    Multi-agent Gaussian Splatting SLAM with loop closure and pose graph fusion — the distributed mapping backbone for autonomous robot teams.