ACC Cloud Simulation Workflow

This project demonstrates a comprehensive workflow for simulating Adaptive Cruise Control (ACC) behavior using real-world driving data in the HighD dataset.

1. Data Upload to Google Cloud Storage (GCS)

• The highd dataset is uploaded to a specified bucket in GCS.
• This dataset includes detailed driving data necessary for ACC simulation.

2. Data Parsing and Filtering with Google Dataflow

• A Dataflow job is created to read the dataset from GCS.
• The job parses the data to identify challenging scenarios based on predefined criteria (e.g. close proximity to another vehicle, high relative velocity, etc.).
• Identified scenarios are stored in a BigQuery table named aac_challenging_scenarios.

3. Publishing Challenging Scenarios to Google Pub/Sub

• The publish_scenarios.py script queries the aac_challenging_scenarios BigQuery table.
• Each scenario is published to the simulate-scenario Pub/Sub topic, triggering downstream processing.
• This step enables asynchronous, event-driven processing of scenarios.

4. Simulating ACC Behavior on Google Compute Engine (GCE)

• A GCE instance is set up with the necessary environment for running the ACC simulation, running the simulate_scenarios.py script.
• The instance subscribes to the simulate-scenario Pub/Sub topic and listens for new challenging scenarios.
• For each received scenario, the ACC behavior is simulated, and the outcome is stored in the simulation_results BigQuery table.

5. Visualization with Looker Studio

• A Looker Studio report is created to visualize the outcomes of the ACC simulations, using the simulation_results table as a data source.
• The report provides insights into how often the ACC system successfully avoided collisions, and other data.
• It allows for easy analysis of ACC performance under different driving conditions.

Looker Studio: https://lookerstudio.google.com/s/lA_BSgXXWsw

Video Demo: https://drive.google.com/file/d/1dpNyOg-aozmjBi-JL6drXkvaOIBctmxN/view?usp=sharing