plozone

3D spatial zone engine — geofencing, octree hole-scanning, realtime sync, voxel pathfinding, and AV sensor fusion. One crate, every domain.

---

Install

[dependencies]
plozone = { version = "0.1", features = ["full"] }

Quick start

use plozone::{
    coord::EnuConverter, octree::OctreeNode,
    store::ZoneStore, zone::{Zone, ZoneEntry},
    scan::{run_scan, ScanMode},
};

// WGS84 origin in Ho Chi Minh City
let conv = EnuConverter::new(10.7626, 106.6601, 0.0);

let store = ZoneStore::from_entries(&[
    ZoneEntry::new(1, Zone::Cylinder {
        center: [10.7626, 106.6601], radius_m: 50.0,
        z_min: 0.0, z_max: 20.0,
    }),
], &conv);

// Is this point inside the zone?
let hits = store.query_geodetic(10.7626, 106.6601, 5.0, &conv);
assert_eq!(hits.as_slice(), &[1]);

// Scan for unsampled regions in the octree
let octree = OctreeNode::new([0.0; 3], 64.0);
let result = run_scan(&octree, &store, 1, &ScanMode::Coarse { depth: 4 });
println!("coverage: {:.1}%", result.coverage_pct);

Why Plozone?

Every spatial project reinvents the same primitives. Plozone answers is this point inside a zone?, did the entity enter or leave?, and is this region fully scanned? across real-world geospatial, game maps, robotics, autonomous vehicles, and IoT — with one consistent API.

Architecture

User Input (WGS84 / game / sensor)    ──►  CoordSystem
                                              │
ZoneStore ◄── R-tree indexed ── Box<dyn ZoneShape>
    │                    │
    ├── query_enu       < 1 µs point-in-zone
    ├── query_geodetic   any lat/lon/alt
    ├── layer queries    filtered by priority
    └── top_priority     tie-breaks on id

OctreeNode ── adaptive subdivision (max depth 17, ~4mm voxels)
    │
    ├── run_scan         single/multiscale hole detection
    ├── range_query      points in bounding box
    └── depth_for_accuracy

ZoneServer ── WebSocket + QUIC + io_uring ── sharded to 300k entities

Two-layer zone model: enum Zone (wire-safe, 6+ built-in shapes, serde) + trait ZoneShape (custom dynamic shapes). Mix both in the same store.

Features

Feature	Provides
(default)	coord, zone, octree, store, scan — zero deps
net	Postcard wire codec, LZ4/ZSTD, dead-reckoning
server	WebSocket server + client, sharding, hysteresis debounce
quic	QUIC transport (quinn + rustls, datagrams)
io_uring	Linux io_uring transport (tokio-uring)
pathfinding	3D voxel A* over octree
game	GameWorld, LayeredMap, ChunkedGameWorld, PortalSystem
terrain	Marching cubes, OBJ/PLY export
av	EKF fusion, sensor FOV, safety envelope, HD map, V2X
bevy	Bevy ECS plugin (Spatial3dPlugin, ZoneTracker)
embedded	no_std linear-scan store
lidar	Pose/quaternion math, scan ingestion
parallel	Rayon-accelerated scanning
geo	geo-types interop
prometheus	Metrics exporter
mimalloc	MiMalloc global allocator

plozone = { version = "0.1", features = ["server", "quic"] }

Transports

Transport	Feature	Latency p99	Use case
WebSocket	server	~5 ms	Browser clients
QUIC	quic	~2 ms	Mobile, 0-RTT reconnect
io_uring	io_uring	< 1 ms	Linux, max throughput

All share the same postcard wire format and ZoneServer core.

Scale

Tier	Entities	Latency p99
Single process	~5k	< 5 ms
Sharded ×16	~80k	< 2 ms
Sharded ×64 + io_uring	~300k	< 1 ms
Multi-process	5M+	< 5 ms

Performance

Operation	Time
1k zone point query	~730 ns
1k octree insert	~435 µs
Hole scan depth 4	~12 µs
WGS84 → ENU	~128 ns
ENU → WGS84	~655 ns

Build & test

cargo build                              # core (zero deps)
cargo test --features full               # 105 tests
cargo bench --features full              # criterion
cargo clippy --features full -- -D warnings

License

MIT. See LICENSE.

Links

• API docs · GitHub · crates.io
• Design document · Contributing