- Gaussian-LIC2 // 2507.04004
- GPU
- WAVE 6
HEIMDALL
THE BRIDGE KEEPER
HEIMDALL fuses LiDAR, camera, and IMU data through continuous-time trajectory optimization using B-spline modeling. It produces photo-realistic Gaussian Splatting maps while maintaining precise 6-DoF localization. The multi-sensor approach handles rapid motion, varying lighting, and geometrically ambiguous environments far better than any single-sensor SLAM system. Multi-sensor fusion is the gold standard for robust SLAM — HEIMDALL combines dense appearance from cameras, precise geometry from LiDAR, and high-frequency motion from IMU to create the most complete spatial representation in the ANIMA stack.
MODULE STATUS: DEVELOPMENTSensors
6DOF
- DIVISION
- ANIMA
- WAVE
- W6
- DOMAIN
- SLAM & 3D
- WAVE 6 // ANIMA SUITE
- FOUNDATION — MULTI-SENSOR FUSION SLAM
NO SINGLE SENSOR IS ENOUGH
Cameras fail in darkness, LiDAR misses color, IMU drifts. Only fusion provides robust perception.
The gold standard for SLAM requires seamless multi-sensor integration.
WHAT HEIMDALL DELIVERS
HEIMDALL fuses LiDAR, camera, and IMU through continuous-time B-spline trajectory optimization producing photo-realistic Gaussian Splatting maps.
CAPABILITIES
- LiDAR + Camera + IMU fusion
- B-spline continuous-time trajectory
- Photo-realistic Gaussian Splatting maps
- Handles rapid motion and varying lighting
WHY THIS IS HARD
Building HEIMDALL requires solving multiple coupled problems:
- 01Synchronizing three sensor modalities with different rates
- 02Continuous-time trajectory modeling
- 03Photo-realistic map generation in real-time
- 04Handling sensor degradation gracefully
HEIMDALL solves these through careful architecture design and rigorous validation.
PROOF, NOT PROMISES
Key metrics:
| METRIC | VALUE |
|---|---|
| Sensors | LiDAR+Cam+IMU |
| Map Type | Gaussian Splatting |
| DoF | 6 |
| Trajectory | Continuous B-spline |
WHAT'S BUILT TODAY
| COMPONENT | STATUS | NOTES |
|---|---|---|
| B-Spline Trajectory | COMPLETE | Continuous-time optimization |
| LiDAR Integration | COMPLETE | Point cloud processing |
| Camera Fusion | IN PROGRESS | Visual feature integration |
| IMU Integration | IN PROGRESS | High-frequency motion |
| Core models | COMPLETE | Fusion SLAM validated |
| API layer | IN PROGRESS | Multi-modal streaming |
WHERE HEIMDALL DEPLOYS
- APP_01
FULL-STACK MAPPING
The most complete spatial representation in ANIMA.
- APP_02
DIGITAL TWINS
Photo-realistic 3D models of real environments.
- APP_03
FIELD OPERATIONS
Robust SLAM in degraded conditions.
UNDER THE HOOD
FOUNDATION: GAUSSIAN-LIC2
- LiDAR + Camera + IMU fusion
- B-spline continuous-time trajectory
- Photo-realistic Gaussian Splatting maps
KEY INNOVATION
HEIMDALL fuses LiDAR, camera, and IMU through continuous-time B-spline trajectory optimization producing photo-realistic Gaussian Splatting maps.
DEPLOYMENT
- REST API
- Docker containerized
- Prometheus metrics
- Configurable backends
COMPUTE
- PRIMARY
- GPU
- EDGE
- Optimized inference
- API
- REST + streaming
PAPERS
- [01]Gaussian-LIC2 (2507.04004)