How zzr005119/carla-extreme-scenario-generator works
This repository is a CARLA-oriented scenario generation and adversarial testing toolkit. A typical workflow creates or selects parameterized scenarios, validates them, runs or registers simulation work, then analyzes risk and feeds results into candidate selection or a scenario library. It also provides conditional-model training, library querying, and a local web interface. The README is empty; relationships below rely on shown excerpts and SOURCE INDEX references, so unsampled implementation details are not inferred.
Main entry points: scenario_features.py Feature encoding · evaluate_adversarial_baselines.py Baseline evaluation · web_task_orchestrator.py Task orchestration · scenario_validator.py Validation and compile
Components 24
Scenario generation
- Feature encoding
core/scenario_features.py - Validation and compile
core/scenario_validator.py - Conditional generators
models/conditional_gmm.py - CVAE training
training/train_cvae.py - Training dataset
training/scenario_dataset.py
Adversarial testing
- Candidate agent
core/adversarial_agent.py - Adversarial loop
core/adversarial_loop.py - Candidate sampling
core/adversarial_sampling.py - Gym environment
core/adversarial_gym_env.py - Proxy executor
core/adversarial_proxy_executor.py
Simulation and evidence
- CARLA scenes
scenes/scene_04_parameterized.py - Sensor writing
core/sensor_pipeline.py - Risk metrics
core/risk_metrics.py
Scenario library
- Scenario library
core/scenario_library.py - Library query
core/scenario_query.py - Candidate scoring
analysis/score_feedback_candidates.py
Interfaces and analysis
- Query command
tools/query_scenario_library.py - Web application
tools/web_app.py - Task orchestration
core/web_task_orchestrator.py - Scenario dashboard
tools/scenario_dashboard.py - Baseline evaluation
tools/evaluate_adversarial_baselines.py - Repeatability analysis
analysis/analyze_carla_repeatability.py
Other components
- Engineer
- CARLA service
Connections 31
- Engineer → Web application: uses
tools/web_app.py - Web application → Scenario dashboard: routes to
tools/web_app.py - Web application → Task orchestration: submits tasks
tools/web_app.py - Task orchestration → Validation and compile: validates
core/web_task_orchestrator.py - Task orchestration → Risk metrics: analyzes risk
core/web_task_orchestrator.py - Task orchestration → CARLA service: registers external run
core/web_task_orchestrator.py - Task orchestration → Scenario library
- Engineer → Query command: queries
tools/query_scenario_library.py - Query command → Library query: matches entries
tools/query_scenario_library.py - Scenario dashboard → Library query: queries
tools/scenario_dashboard.py - Scenario dashboard → Query command: uses
tools/scenario_dashboard.py - Scenario library → Feature encoding: uses encoding
core/scenario_library.py - Scenario library → Validation and compile: validates records
core/scenario_library.py - Candidate scoring → Feature encoding: uses features
analysis/score_feedback_candidates.py - Candidate scoring → Validation and compile: validates candidates
analysis/score_feedback_candidates.py - Conditional generators → Feature encoding: encodes records
models/conditional_gmm.py - CVAE training → Training dataset: loads arrays
training/train_cvae.py - CVAE training → Conditional generators: trains model
training/train_cvae.py - CVAE training → Feature encoding: uses feature schema
training/train_cvae.py - Training dataset → Feature encoding: converts records
training/scenario_dataset.py - Candidate agent → Feature encoding: uses features
core/adversarial_agent.py - Candidate agent → Validation and compile: validates candidates
core/adversarial_agent.py - Baseline evaluation → Candidate agent: evaluates
tools/evaluate_adversarial_baselines.py - Baseline evaluation → Adversarial loop: runs loop
tools/evaluate_adversarial_baselines.py - Baseline evaluation → Candidate sampling: samples candidates
tools/evaluate_adversarial_baselines.py - Baseline evaluation → Feature encoding: uses features
tools/evaluate_adversarial_baselines.py - Baseline evaluation → Validation and compile: validates scenarios
tools/evaluate_adversarial_baselines.py - CARLA scenes → Risk metrics: uses risk metrics
scenes/scene_04_parameterized.py - CARLA scenes → Sensor writing: writes sensor data
scenes/scene_04_parameterized.py - CARLA service → CARLA scenes: simulates
- Repeatability analysis → CARLA service: analyzes run evidence