Skip to content

How to Run

To run this project, you need to provide the analysis base configuration and log information at runtime. Their structure is as follows.

Base Configuration

The base configuration defines the parameters required during the analysis process, including the time window for analysis and the specific settings for each module, as shown below:

🔧 Example Configuration (config/RCanaylsis.yaml)

log_dir: /path/to/logs
effect_time_window: 10
display_results: true
module_setting:
perception:
    perc_score_threshold: 0.5
prediction:
    prediction_window: 10
    pred_score_threshold: 0.1
planning:
    planning_toleration_future_step: 5
    planning_toleration: 1

Explanation of Configuration Fields

📝 Explanation of Configuration Fields

  • log_dir: Path to the input log directory.
  • effect_time_window: Time window (in frames) to consider for root cause analysis.

  • display_results: Whether display results on screen.

  • module_setting:

    • perception.perc_score_threshold: Minimum confidence score for the perception module to consider a result valid.
    • prediction.prediction_window: Time horizon (in frames) used by the prediction module(If it differs from the settings of the running prediction module, the smaller value between the two will be used by default.).
    • prediction.pred_score_threshold: Confidence threshold for predictions to be considered reliable.
    • planning.planning_toleration_future_step: The planner allows a small tolerance for errors after a number of future steps.
    • planning.planning_toleration: Tolerance value for planning errors (in meters).

Custom Configuration

If you need to set the analysis configurations, there are two ways to do so:

  1. Directly modify the project's default configuration file. {project}/ADSRootCauseAnalysis/config/RCanalysis.yaml

  2. (Recommended) Create a custom configuration file and specify it at runtime.

You can create a new configuration file with same format, and running with

python main.py --config-path <your/file/path> --config-name <your/file/name>

Input Data

The input data of ADS Root Cause Anaylsis is the running log of ADS, the details can be found in Input Format.

Integration Supports

This project can be run as a standalone application or imported as a Python package for integration into other projects.

In summary, RCHandler wraps all core functionalities and enables standalone execution by importing and using this class directly. To run this package, you need to create a configuration file as an OmegaConf DictConfig and pass it as a parameter to initialize the RCHandler.

from ADSRootCauseAnalysis import RCHandler
from omegaconf import DictConfig
from hydra import initialize, compose

... your code

with initialize(config_path="ADSRootCauseAnaylsis/config"):
    cfg = compose(config_name="RCanaylsis")
    handler = RCHandler(cfg)

... follow code