Disciplines · Proposals

Oya Domain Proposal: AI-Controlled Drone & Swarm Management Platform

Oya is proposed as a new platform capability domain for the Oshun monorepo, providing comprehensive AI-controlled drone and swarm management capabilities.

13sections9 minread

On this page

Proposed Domain Name: Oya (Yoruba goddess of wind, storms, and transformation) Scope: scope:oya Type: Platform Capability Domain Date: January 2026


Executive Summary#

Oya is proposed as a new platform capability domain for the Oshun monorepo, providing comprehensive AI-controlled drone and swarm management capabilities. Like Aja (motion intelligence), Bellona (engine bridges), and Isis (generative factory), Oya would be consumed by product domains (Lilith, Yemaya, Aphrodite) rather than being a standalone product itself.

The domain encompasses:

  • Multi-drone coordination and swarm intelligence
  • AI-powered autonomous cinematography
  • Indoor/outdoor navigation systems
  • Real-time pose estimation integration for human tracking
  • Hardware abstraction for multiple drone platforms
  • Sensor fusion and edge AI processing

1. Domain Name Rationale#

Oya (Òyá)#

In Yoruba mythology, Oya is the powerful goddess of wind, storms, lightning, and transformation. She is one of the most formidable orishas, associated with:

  • Wind and Air - Commands the winds, from gentle breezes to fierce storms
  • Lightning and Thunder - Swift, powerful, and precise
  • Transformation - Goddess of change and adaptation
  • Guardian of the Dead - Protector during transitions

Why Oya fits:

  • Thematically aligned with flight - As goddess of wind, she embodies aerial movement
  • Yoruba mythology consistency - Matches existing Oshun domains (Yemaya, Aja are also Yoruba)
  • Swift and adaptive - Perfect for AI-powered autonomous systems
  • Transformation aspect - Reflects the transformative nature of drone technology
  • Storm coordination - Multiple forces working together, like drone swarms
  • Female deity - Maintains the naming convention of Oshun domains

2. Core Capabilities#

2.1 Flight Control & Navigation#

Capability Technology Description
GPS Navigation RTK-GPS, GNSS multi-constellation Centimeter-level outdoor positioning
Indoor Positioning UWB (10cm accuracy), VIO, SLAM GPS-denied environment navigation
Visual SLAM ORB-SLAM3, Isaac ROS VSLAM Real-time 3D mapping and localization
Obstacle Avoidance ToF sensors, stereo depth, AI vision 360-degree collision prevention
Path Planning A*, RRT, MPC-based trajectories Optimal route computation

2.2 Swarm Coordination#

Capability Algorithm Description
Collision Avoidance ORCA (Optimal Reciprocal Collision Avoidance) Multi-agent collision-free motion
Formation Control Reynolds flocking (separation, alignment, cohesion) Coordinated swarm behavior
Task Allocation SARG, Hybrid DRL + Distributed Optimization Optimal mission assignment
Communication MAVLink protocol, mesh networking Inter-drone data exchange
Consensus Federated learning architecture Distributed decision making

2.3 AI Cinematography#

Capability Technology Description
Subject Tracking YOLO11-Pose, Skydio Shadow Real-time human/object following
Shot Composition CineMPC, rule of thirds Autonomous framing decisions
Multi-Camera Coordination MAPF algorithms Multiple drone viewpoint management
Cinematic Path Planning B-spline/Bezier trajectories Smooth, professional camera moves
KeyFrame Animation AI-generated transitions Complex camera path automation

2.4 Sensor Integration#

Sensor Type Examples Use Cases
RGB Cameras 4K-8K, 100MP sensors Video capture, photography
Thermal/IR FLIR Boson+, radiometric Night operations, inspection
Depth Sensors Intel RealSense, ToF Obstacle detection, 3D mapping
LiDAR Ouster OS0, Velodyne High-precision mapping
Audio Beamforming arrays, speakers Two-way communication
IMU/GPS 9-axis, multi-constellation Positioning, orientation

2.5 Pose Estimation Integration#

Capability Algorithm Description
2D Pose Detection MediaPipe, YOLO11-Pose, RTMPose Real-time keypoint extraction
3D Reconstruction SMPL, SKEL, HMR2.0 Full body mesh recovery
Multi-View Fusion SelfPose3d, voxel-based methods Multi-drone 3D analysis
Form Analysis Joint angle calculation Fitness/yoga/sports coaching
Gait Analysis 3DGait, spatiotemporal metrics Movement pattern analysis

2.6 Edge AI Processing#

Platform Performance Use Case
NVIDIA Jetson Orin 275 TOPS High-end autonomous operations
NVIDIA Jetson Thor 2,070 TFLOPS Next-gen AI processing
Qualcomm QRB5165 15 TOPS 5G-connected drones
ModalAI VOXL 2 15+ TOPS, 16g Lightweight autonomous flight

3. Library Architecture#

text
libs/oya/
├── core/
│   ├── types/                    # Core type definitions
│   ├── constants/                # Physical constants, limits
│   ├── utils/                    # Common utilities
│   └── errors/                   # Domain-specific errors
│
├── flight-control/
│   ├── kinematics/               # Drone physics, dynamics
│   ├── attitude-control/         # Orientation management
│   ├── velocity-control/         # Speed regulation
│   └── mission-execution/        # Waypoint following
│
├── navigation/
│   ├── gps/                      # GNSS positioning
│   ├── indoor-positioning/       # UWB, VIO, motion capture
│   ├── visual-slam/              # ORB-SLAM, RTAB-Map integration
│   ├── path-planning/            # A*, RRT, trajectory optimization
│   └── obstacle-avoidance/       # Sensor fusion, collision prevention
│
├── swarm/
│   ├── coordination/             # Multi-agent coordination
│   ├── formation/                # Formation flying algorithms
│   ├── task-allocation/          # Mission distribution
│   ├── consensus/                # Distributed agreement
│   └── communication/            # Inter-drone messaging
│
├── cinematography/
│   ├── shot-planning/            # Composition algorithms
│   ├── subject-tracking/         # Person/object following
│   ├── trajectory-optimization/  # Smooth path generation
│   ├── multi-camera/             # Multi-drone coordination
│   └── streaming/                # Real-time video streaming
│
├── sensors/
│   ├── camera/                   # RGB, thermal, depth cameras
│   ├── lidar/                    # LiDAR integration
│   ├── imu/                      # Inertial measurement
│   ├── gps/                      # GNSS receivers
│   ├── audio/                    # Microphones, speakers
│   └── fusion/                   # Multi-sensor fusion
│
├── pose-estimation/
│   ├── detection/                # 2D pose detection
│   ├── reconstruction/           # 3D body reconstruction
│   ├── multi-view/               # Multi-camera fusion
│   ├── tracking/                 # Temporal tracking
│   └── analysis/                 # Form analysis, metrics
│
├── hardware-abstraction/
│   ├── dji/                      # DJI MSDK, OSDK, PSDK adapters
│   ├── px4/                      # PX4 autopilot integration
│   ├── ardupilot/                # ArduPilot integration
│   ├── mavlink/                  # MAVLink protocol
│   └── generic/                  # Abstract interfaces
│
├── edge-ai/
│   ├── inference/                # Model inference runtime
│   ├── optimization/             # Quantization, pruning
│   ├── models/                   # Pre-trained models
│   └── deployment/               # Edge deployment utilities
│
├── safety/
│   ├── geofencing/               # Virtual boundaries
│   ├── emergency/                # Failsafe procedures
│   ├── human-aware/              # Social navigation
│   └── compliance/               # Regulatory compliance
│
├── telemetry/
│   ├── collection/               # Real-time data gathering
│   ├── streaming/                # Live telemetry broadcast
│   ├── logging/                  # Flight data recording
│   └── analysis/                 # Post-flight analysis
│
├── simulation/
│   ├── gazebo/                   # Gazebo integration
│   ├── airsim/                   # AirSim integration
│   ├── jmavsim/                  # jMAVSim integration
│   └── physics/                  # Physics simulation
│
├── client/                       # TypeScript SDK for consumers
├── database/                     # Flight logs, telemetry storage
└── event-handlers/               # Event bus integration

Estimated Library Count: 45-55 libraries


4. Integration with Existing Domains#

4.1 Integration Matrix#

Domain Oya Provides Domain Provides Integration Pattern
Aja Drone camera feeds Pose estimation, motion capture Oya sends video → Aja processes → returns skeleton data
Aphrodite Multi-angle capture Streaming infrastructure Oya provides camera positions → Aphrodite streams
Lilith Meditation flyovers Experience orchestration Lilith triggers missions → Oya executes
Yemaya Cinematography shots Scene composition Yemaya defines shots → Oya plans paths
Isis Raw video feeds AI enhancement, thumbnails Oya captures → Isis processes
Sophia Flight data Learning, recommendations Oya logs → Sophia optimizes
Bellona Real-time position Engine visualization Oya publishes → Bellona renders

4.2 Detailed Integration Scenarios#

Aphrodite + Oya: Multi-Angle Performer Capture#

text
Aphrodite (Live Streaming Platform)
        │
        ├── Requests: "Capture performer from 4 angles"
        │
        ▼
Oya (Drone Swarm)
        │
        ├── Coordinates 4 drones in formation
        ├── Tracks performer using pose estimation
        ├── Maintains cinematic framing
        │
        ▼
    ┌───┴───┐
    │       │
    ▼       ▼
  Aja     Aphrodite
(Pose)   (Streaming)

Lilith + Oya: Fitness Form Analysis#

text
Lilith (Wellness Platform)
        │
        ├── User starts yoga session
        │
        ▼
Oya (Drone Controller)
        │
        ├── Positions drone for optimal view
        ├── Captures multi-angle video
        │
        ▼
Aja (Motion Analysis)
        │
        ├── Extracts 33 body keypoints
        ├── Calculates joint angles
        │
        ▼
Lilith (Feedback)
        │
        └── Provides real-time form corrections

Yemaya + Oya: Cinematic Production#

text
Yemaya (Production Studio)
        │
        ├── Director specifies shot sequence
        │
        ▼
Oya (Cinematography)
        │
        ├── Plans optimal camera paths
        ├── Executes with B-spline trajectories
        ├── Coordinates multi-drone shots
        │
        ▼
    ┌───┴───┐
    │       │
    ▼       ▼
  Isis    Bellona
(Process) (Preview)

4.3 Dependency Rules#

text
# Oya CAN depend on:
- @oshun/* (shared foundation)
- @aja/* (motion capture, pose estimation)
- @sophia/* (learning, recommendations)

# Oya CANNOT depend on:
- @lilith/* (product domain)
- @yemaya/* (product domain)
- @aphrodite/* (product domain)

# Products CAN consume Oya via:
- gRPC/REST APIs
- Event bus subscriptions
- TypeScript client SDK

5. Hardware Platform Support#

5.1 Consumer/Prosumer Drones#

Platform Integration SDK Capabilities
DJI Mavic 4 Pro Full MSDK V5 100MP, 6K video, 51min flight
DJI Mini 5 Pro Partial MSDK V5 LiDAR, <250g, 45min flight
Skydio X10 Full Skydio SDK NightSense, FLIR thermal
Autel EVO II Full Autel SDK 6K video, 40min flight

5.2 Enterprise/Professional Drones#

Platform Integration SDK Capabilities
DJI Inspire 3 Full OSDK 8K cinema, full-frame
DJI Matrice 400 Full OSDK/PSDK 59min, LiDAR payloads
Freefly Alta X Full PX4 35lb payload, cinema cameras
Skydio X10D Full Skydio SDK NDAA compliant, government

5.3 Indoor/Development Drones#

Platform Integration SDK Capabilities
ModalAI Starling 2 Full PX4/MAVSDK VIO, SLAM, 280g
Crazyflie Full cflib 27g, research, swarms
HOVERAir X1 Partial - 125g, autonomous selfie
Ring Always Home Monitor - Indoor security

5.4 Custom/Open-Source#

Platform Integration SDK Capabilities
PX4-based Full MAVSDK, ROS 2 Customizable
ArduPilot-based Full DroneKit, MAVLink Community support
Pixhawk hardware Full MAVLink Industry standard

6. Use Cases by Oshun Domain#

6.1 Aphrodite Integration#

Use Case Description Drones Key Features
Multi-Angle Performance Capture Stream performer from multiple simultaneous angles 2-8 Synchronized video, subject tracking
360-Degree Environment Create immersive viewing experience 4+ Formation flying, coordinated cameras
Interactive Drone Responses Drones react to tips/engagement 1-4 Event-driven commands, LED effects
VR/AR Capture Stereo capture for VR viewing 2 Synchronized stereo, depth data

6.2 Lilith Integration (Wellness/Meditation)#

Use Case Description Drones Key Features
Yoga/Fitness Form Checking Real-time posture feedback 1-3 Pose estimation, multi-view analysis
Workout Recording Automatic exercise documentation 1-2 Subject tracking, shot variety
Mirror Replacement Drone with display shows user 1 Screen display, positioning
Movement Coaching AI-powered form correction 2-4 Multi-angle analysis, real-time feedback

6.3 Yemaya Integration (Production)#

Use Case Description Drones Key Features
Cinematic Capture Professional film-quality shots 1-4 8K cameras, cinematic paths
Virtual Production Real-time tracked shots 1-2 RTK positioning, Unreal integration
Documentary Coverage Multi-subject tracking 4-8 AI subject selection, formation
Motion Reference Actor movement capture 2-6 Multi-view pose estimation

6.4 Fashion/Styling Application#

Use Case Description Drones Key Features
Wardrobe Scanning 3D closet inventory 1 RGB-D scanning, cataloging
Full-Body Assessment 360-degree body measurement 2-4 Photogrammetry, AI measurement
Outfit Preview Mirror drone for self-viewing 1 Display screen, tracking
Try-On Assistance Real-time fit feedback 1-2 Pose estimation, overlay

6.5 Novel Applications#

Use Case Description Drones Key Features
Indoor Security Patrol Automated home monitoring 1 Preset paths, anomaly detection
Telepresence Flying video call platform 1 2-way audio/video, following
Plant Care Garden monitoring and care 1-2 NDVI imaging, thermal
Pet Tracking Follow and record pets 1 Animal tracking, safe operation
Light Shows Synchronized LED displays 10-100+ Formation, LED control

7. State-of-the-Art Technical Features#

7.1 Swarm Intelligence#

Feature Algorithm Performance
ORCA Collision Avoidance Optimal Reciprocal Collision Avoidance <10ms planning, 100+ agents
EN-MASCA Patrolling Enhanced Multi-Agent Swarm Control 96% area coverage
DRL Task Allocation Hybrid Deep RL + Distributed Optimization 27.8% better than standalone
SARG Framework Swarm Allocation and Route Generation Real-time 3D trajectory planning
MAVLink Protocol Lightweight binary telemetry 255 concurrent systems, 14 bytes overhead

7.2 AI Cinematography#

Feature Technology Capability
CineMPC Model Predictive Control Intrinsic + extrinsic camera control
Subject Tracking YOLO11-Pose + Kalman Filter 30+ FPS, occlusion recovery
KeyFrame Animation AI-generated transitions Complex multi-keyframe paths
Shot Composition Rule of thirds, golden ratio Automatic framing
Multi-Drone Coordination MAPF algorithms Field-tested with 3+ UAVs

7.3 Indoor Navigation#

Feature Technology Accuracy
UWB Positioning LinkTrack, Pozyx 10cm (2D), 30cm (3D)
Visual-Inertial Odometry ORB-SLAM3, VINS-Mono 0.08m final pose
Motion Capture Integration OptiTrack, Vicon <0.3mm positional
Event-Based VIO Event cameras + IMU 50% error reduction
Depth Sensing Intel RealSense D435i Real-time obstacle detection

7.4 Pose Estimation#

Feature Model Performance
Real-Time Detection RTMPose-m 75.8% AP @ 90+ FPS
Aerial HPE FlyPose (ViTPose-S) Up to 40m altitude
3D Reconstruction HSMR + SKEL 10mm+ improvement over HMR2.0
Multi-View Fusion SelfPose3d Self-supervised, no 3D labels
Edge Inference Jetson Orin + TensorRT 30+ FPS onboard

7.5 Hardware Capabilities#

Feature Platform Specification
Camera Resolution DJI Mavic 4 Pro 100MP, 6K@60fps
Thermal Imaging FLIR Boson+ 640x512, 30mK sensitivity
Flight Time DJI Mavic 4 Pro 51 minutes
Edge AI NVIDIA Jetson Orin 275 TOPS
NightSense Skydio X10 Zero-light autonomous operation

7.6 Communication & Streaming#

Feature Technology Specification
Video Transmission DJI O4+ 30km range, 10-bit HDR
Low-Latency Streaming WebRTC <150ms latency
Mesh Networking Meshmerize 100+ nodes, <100ms latency
5G Integration Skydio Connect Unlimited range with cellular
Multi-Camera Sync Timecode + genlock Sub-frame synchronization

8. Safety & Compliance#

8.1 Safety Features#

Feature Description
Propeller Guards Mandatory for indoor use, full enclosure preferred
Speed Limiting Configurable max speed (2-5 mph indoor)
Geofencing Static, dynamic, and custom virtual boundaries
Emergency Landing Automatic failsafe procedures
Low Battery RTH Return-to-home on battery threshold
Obstacle Avoidance 360-degree sensor coverage
Human-Aware Navigation Social navigation algorithms

8.2 Noise Reduction#

Technique Reduction Technology
Serrated Edges ~5.7% decrease Micro-vortex generation
Toroidal Propellers 47% quieter No tip vortex
Lower RPM Significant Larger props, slower spin
Target <60 dB EU urban limit compliance

8.3 Regulatory Compliance#

Region Regulation Oya Support
USA FAA Part 107 Remote ID, registration tracking
EU EASA regulations 60 dB limit compliance
NDAA US Government Skydio X10D integration
Indoor Not FAA regulated Property owner rules support

9. Development Priorities#

Phase 1: Core Infrastructure (Months 1-3)#

  • Core types and constants
  • MAVLink protocol implementation
  • PX4/ArduPilot adapters
  • Basic telemetry collection
  • TypeScript client SDK

Phase 2: Navigation & Control (Months 4-6)#

  • GPS navigation
  • Indoor positioning (UWB, VIO)
  • Basic obstacle avoidance
  • Path planning algorithms
  • Mission execution

Phase 3: Swarm Capabilities (Months 7-9)#

  • Multi-drone coordination
  • ORCA collision avoidance
  • Formation flying
  • Task allocation
  • Mesh communication

Phase 4: AI Cinematography (Months 10-12)#

  • Subject tracking
  • Shot composition
  • Trajectory optimization
  • Multi-camera coordination
  • Streaming integration

Phase 5: Pose Estimation Integration (Months 13-15)#

  • 2D pose detection pipeline
  • 3D reconstruction
  • Multi-view fusion
  • Form analysis algorithms
  • Aja domain integration

Phase 6: Domain Integrations (Months 16-18)#

  • Aphrodite integration
  • Lilith integration
  • Yemaya integration
  • Production hardening
  • Documentation

10. Technology Stack Summary#

Languages & Frameworks#

  • TypeScript: Client SDK, event handlers
  • Python: ML/AI pipelines, pose estimation
  • Rust: Performance-critical flight control
  • C++: Edge AI inference, hardware interfaces

Key Dependencies#

  • MAVLink/MAVSDK: Flight control protocol
  • ROS 2: Robotics middleware
  • PX4/ArduPilot: Autopilot systems
  • NVIDIA TensorRT: Edge AI inference
  • OpenCV: Computer vision
  • MediaPipe/YOLO: Pose estimation

Infrastructure#

  • Redis Streams: Event messaging
  • PostgreSQL: Flight logs, telemetry
  • MinIO: Video/image storage
  • Gazebo/AirSim: Simulation

11. Success Metrics#

Metric Target Measurement
Indoor Positioning <15cm accuracy UWB + VIO fusion
Swarm Coordination 10+ drones Collision-free operation
Video Latency <200ms WebRTC streaming
Pose Estimation 30+ FPS On-device inference
Battery Efficiency >40 min flight Optimized paths
Noise Level <60 dB EU compliance

Conclusion#

Oya represents a comprehensive platform capability domain that would enable sophisticated drone-based features across the entire Oshun ecosystem. By following the established architectural patterns (capability domain consumed by product domains), it provides maximum reusability while maintaining clear boundaries.

The domain name Oya (Yoruba goddess of wind, storms, and transformation) maintains thematic consistency with Oshun's mythology-based naming (alongside Yemaya and Aja) while perfectly capturing the essence of flight, swift coordination, and the transformative power of drone swarms working in harmony.

Key differentiators:

  1. Multi-platform support: DJI, PX4, ArduPilot, Skydio
  2. Indoor-first: UWB, VIO, SLAM for GPS-denied operation
  3. Swarm-native: Built for multi-drone coordination from the start
  4. AI-powered: Cinematography, pose estimation, navigation
  5. Integration-ready: Designed for Oshun domain consumption