Upload README.md with huggingface_hub

README.md

@@ -0,0 +1,67 @@
+---
+license: cc-by-nc-4.0
+task_categories:
+- other
+tags:
+- synthetic-data
+- geospatial
+- gps
+- trajectory
+- logistics
+size_categories:
+- 10M<n<100M
+---
+
+# FreeSyntheticGPSTrajectories100M
+
+100 million fully synthetic GPS trip points across roughly 944,000 trips, built for developers and researchers working on logistics, rideshare, fleet-tracking, or mapping systems who need realistic-shaped trajectory data without touching real location data. Every trip is a continuous, physically plausible path — not independent random points — making this suitable for testing route reconstruction, trip segmentation, geofencing, and time-series/spatial pipelines. No real vehicles, devices, or individuals are represented; every trip is randomly generated.
+
+## Schema
+
+| Column | Type | Description |
+|---|---|---|
+| point_id | string | Unique identifier for this GPS ping |
+| trip_id | string | Groups points belonging to the same continuous trip |
+| vehicle_id | string | Vehicle/device identifier, reused across multiple trips |
+| trip_sequence_num | int32 | Position of this point within its trip (0-indexed) |
+| timestamp | string | ISO 8601 UTC timestamp |
+| latitude | float32 | |
+| longitude | float32 | |
+| speed_kmh | float32 | |
+| heading_deg | float32 | Compass heading, 0-360 |
+| accuracy_m | float32 | Simulated GPS accuracy in meters |
+
+## Format
+
+Single Parquet file, Snappy compression, ~3.2 GB, 100,000,000 rows.
+
+## Quick start
+
+```python
+import duckdb
+duckdb.sql("SELECT * FROM 'geo_gps_100M.parquet' LIMIT 10").show()
+```
+
+```python
+import pyarrow.parquet as pq
+pf = pq.ParquetFile("geo_gps_100M.parquet")
+for batch in pf.iter_batches(batch_size=100000):
+    ...  # process each batch
+```
+
+```python
+from datasets import load_dataset
+ds = load_dataset("ziadatalabs/FreeSyntheticGPSTrajectories100M", streaming=True)
+```
+
+## Notes
+
+Points are ~10 seconds apart within a trip. Trips originate in one of 35 major world cities (weighted toward larger metros) and follow a continuous random-walk path — heading drifts gradually and speed follows realistic urban/highway/stop patterns rather than teleporting between unrelated coordinates. Vehicles are reused across multiple trips to mirror a fleet-style usage pattern. All entirely synthetic — no real GPS traces, devices, or individuals are represented.
+
+## License & Usage
+
+Licensed under CC-BY-NC-4.0. Free for personal, research, and educational use with attribution. Not licensed for commercial use.
+
+---
+
+Created by Zia Data Labs. Questions or feedback: zia.data.team@protonmail.com