Skip to main content

Token-level PDMS trajectory scoring utilities for HelloData.

Project description

Trajectory Scorer

HelloData 轨迹的 token-level PDMS 评分与可视化工具。

当前流程一次评估一个 token/window。clip 只是帧数据来源;评测入口消费一个 ScoringContext window 和一条候选轨迹。

安装

在仓库根目录执行:

pip install .

开发模式安装(代码改动可立即生效):

pip install -e ".[dev]"

仓库结构

  • traj_scorer/:评分库代码。
  • traj_scorer/pdms.pyevaluate_context() 和 PDMS 聚合逻辑。
  • traj_scorer/metric_*.py:各个子指标实现。
  • traj_scorer/hellodata.py:HelloDataImporter 适配层。
  • scripts/repair_token_level_clip_3_155.py:当前 token-level 验证与 BEV 可视化入口。
  • tests/:聚焦指标行为的回归测试。
  • docs/pdms_metrics.md:子指标定义与聚合公式说明。

主 Token-Level 命令

示例:clip 3_155、token 230、随机 50 条轨迹库候选。

python scripts/repair_token_level_clip_3_155.py \
  --output-dir output/token_level_repair_3_155_token230 \
  --library-count 50 \
  --token-starts 230 \
  --token-length 40 \
  --random-seed 155

输出包括:

  • expert_sanity.json/csv/md
  • library_perturbation_comparison.json/csv
  • bev_by_metric/*.png

BEV 图会在 token 起点自车坐标系下绘制 OCC freespace、LD、当前/未来 ego、 当前/未来 agents,以及按指标分数着色的候选轨迹。

指标

详细说明见 docs/pdms_metrics.md

已实现:

  • NC:无责碰撞。
  • DAC:可行驶区域合规,优先使用 OCC freespace。
  • EP:自车进度。
  • TTC:短时投影碰撞检查。
  • LK:基于当前自车中心线的车道保持。
  • HC:历史舒适性。

暂时默认通过:

  • DDC:行驶方向合规。
  • TLC:红绿灯合规。

验证

运行:

python -m py_compile scripts/repair_token_level_clip_3_155.py traj_scorer/*.py
python -m pytest -q

当前测试覆盖 EP 进度排序和 LK 中心线行为,包括“相邻中心线不应掩盖偏离当前车道”的情况。

打包发布给他人

先构建 wheel/sdist:

python -m build

构建产物会在 dist/ 下,别人可直接安装:

pip install dist/traj_scorer-0.1.0-py3-none-any.whl

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

traj_scorer-0.1.0.tar.gz (17.1 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

traj_scorer-0.1.0-py3-none-any.whl (20.6 kB view details)

Uploaded Python 3

File details

Details for the file traj_scorer-0.1.0.tar.gz.

File metadata

  • Download URL: traj_scorer-0.1.0.tar.gz
  • Upload date:
  • Size: 17.1 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.2.0 CPython/3.13.12

File hashes

Hashes for traj_scorer-0.1.0.tar.gz
Algorithm Hash digest
SHA256 52569dd91a472f507e4bd8e040127ee1bc4feec800b8a4bd7b717f4421af76cc
MD5 914b6c29c3871412255528f5ff0ea6ec
BLAKE2b-256 5f28671d9d23f533df9eb9c78c82829f16fe381539a6ec285fada1b0d02eb6a6

See more details on using hashes here.

File details

Details for the file traj_scorer-0.1.0-py3-none-any.whl.

File metadata

  • Download URL: traj_scorer-0.1.0-py3-none-any.whl
  • Upload date:
  • Size: 20.6 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.2.0 CPython/3.13.12

File hashes

Hashes for traj_scorer-0.1.0-py3-none-any.whl
Algorithm Hash digest
SHA256 fcbe8a883bd5ee83e698939682d9d9c2659029f74ea91cc6977ab0cb0876ac2e
MD5 ef831bcec63dc5d08c726c98f5bb6a0f
BLAKE2b-256 112e5ac35dce189cb7a82fb42936863077d9e4f1b03b7d6b46c04e09861f8339

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page