# map

**Repository Path**: REIGE/map

## Basic Information

- **Project Name**: map
- **Description**: x先谢谢谢谢谢谢谢谢
- **Primary Language**: Unknown
- **License**: Not specified
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2025-09-20
- **Last Updated**: 2025-12-02

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

Offline HMM Map-Matching (Cellular-Aided)

Overview
- Implements offline map-matching using a Hidden Markov Model and Viterbi decoding.
- Emission uses spatial distance to roads plus cellular features: TA (timing advance) and RSRP, and serving sector alignment.
- Transition uses a kinematic speed prior and penalizes frequent road switching.

Data Formats (CSV)
1) Observations (columns; Chinese names auto-mapped):
- 时间戳 (timestamp): int seconds
- 经度 (lon): float
- 纬度 (lat): float
- 基站id (station_id): optional string
- 主服小区id (primary_cell_id): optional string
- TAValue (ta_value): optional float
- 邻区数量 (neighbor_count): optional int
- gnodebid: optional int
- cellid: optional int
- rsrp: optional float (dBm)

2) Road vectors:
- 道路id (road_id): string
- 经度 (lon): float
- 纬度 (lat): float
- seq: optional int ordering per road_id (auto-generated if missing)

3) Base stations:
- gnodebid: int
- cellid: int
- 经度 (lon): float
- 纬度 (lat): float
- 方位角 (azimuth_deg): optional float degrees

Install
Use Poetry (or adapt to pip):
```bash
cd /Users/reige/Downloads/map
poetry install
```

Sample Data
- Provided under `data/`:
  - `data/roads.csv` 两条相交道路（横向 R1、纵向 R2）
  - `data/stations.csv` 一个站点三扇区（cellid 1/2/3）
  - `data/obs.csv` 一小段沿 R1 行进并跨越 R2 的观测序列

Run (sample)
```bash
poetry run python -m mapmatch.cli \
  --obs data/obs.csv \
  --roads data/roads.csv \
  --stations data/stations.csv \
  --out data/corrected.csv \
  --radius 60 --maxc 8 --sigma 30 --ta_unit_m 78 \
  --speed 13.9 --speed_sigma 7 --json_out
```

Output
- corrected CSV: timestamp, road_id, proj_x, proj_y (proj in EPSG set by --epsg, default 3857)
- optional JSON: log_likelihood, num_observations

Notes
- Candidate search uses a simple brute-force over all road geometries. For large networks, replace with STRtree spatial index.
- TA meter per unit depends on RAT/vendor; adjust via --ta_unit_m.
- Transition can be improved using actual network path length along roads.