CARLA ? C-Shenron based Simualtor for Sensor data generation.

9.6 KiB

Raw Blame History

Fox CARLA ADAS Simulation Project: Agent Intelligence (gemini.md)

This document is the consolidated source of truth for AI agents working on the Fox CARLA Simulation project. It combines high-level architecture, component-level details, and best practices.

🏗️ 1. Project Overview & Architecture

Objective: A modular, scenario-driven simulation framework built on CARLA 0.9.16 for demonstrating structured ADAS driving scenarios with multi-modal sensor data.

The Pipeline: CARLA Simulator → Multi-Sensor Capture → Dataset (PNG/NPY/JSONL) → [SHENRON SYNTHESIS] → MCAP Conversion → Foxglove Visualization

Repository Layout

dashboard.bat: One-click launcher for the Flask GUI orchestrator.
run.bat: One-click runner for the simulation (activates carla312 conda env).
config.py: Global simulation constants (FPS, sensor specs, ego model).
src/main.py: Thin CLI wrapper for the PipelineManager.
src/pipeline/: Core stage-based architecture (Manager + Stages).
src/pipeline/base.py: PipelineContext dataclass and PipelineStage ABC.
src/pipeline/manager.py: Sequential stage executor with cleanup guarantees.
src/pipeline/stages/: Individual stage implementations.
src/processing/: Centralized physics and sensor augmentation logic.
src/processing/physics.py: Radial velocity injection, ADAS metrics, actor classification.
src/sensors.py: SensorManager — Handles camera, radar, and lidar sync.
src/recorder.py: Recorder — Serializes raw frames to disk asynchronously.
scenarios/: Contains scenario implementations (e.g., braking.py, cutin.py, showcase.py).
scripts/data_to_mcap.py: Production MCAP converter with full metrology suite.
scripts/generate_shenron.py: Orchestrates Shenron radar synthesis per session.
scripts/test_shenron.py: Standalone diagnostic testbench for radar iteration.
scripts/ISOLATE/: High-fidelity radar simulation engine.
scripts/ISOLATE/sim_radar_utils/: Visualization and DSP utilities.
scripts/ISOLATE/sim_radar_utils/plots.py: Single source of truth for all radar heatmap rendering (FastHeatmapEngine, postprocess_ra, scan_convert_ra).
scripts/ISOLATE/sim_radar_utils/config.yaml: DSP processor and plot limit configuration.
dashboard/: Flask backend and static web assets for the GUI.
intel/: Detailed deep-dive documentation for specific components.
intel/CHRONICLES.md: Project History — Weekly evolution and major milestones.
intel/memory_update.md: Agent Protocol (v1.1) — Standardized guide for repo memory updates.
intel/internal/[version]/: Versioned release changelogs and walkthroughs.
intel/radar/: Deep dives into radar physics (e.g. research/isotropic_illumination_problem.md).

🛠️ 2. Component Reference

`src/pipeline/manager.py` (The Director)

Role: Sequential executor for pipeline stages.
Workflow: Reads PipelineContext → Skips stages if requested → Executes Sim → Shenron → MCAP → Video.
Error Handling: Halts the pipeline on critical stage failures but ensures cleanup() is called for all started stages (reverse order).

`src/main.py` (CLI Entry Point)

Role: Parses arguments and initializes the PipelineManager.
Features: Supports selective execution via --only-mcap, --only-shenron, --skip-shenron, --skip-mcap, --skip-sim.
Session Reuse: --session <path> allows re-processing existing data without re-running the simulation.

`src/pipeline/stages/` (The Workers)

SimulationStage (sim_stage.py): Handles CARLA connection, ego spawning, sensor capture loop, and stop.flag detection.
ShenronStage (shenron_stage.py): Triggers physics-based radar synthesis via generate_shenron.py. Non-fatal on error.
McapStage (mcap_stage.py): Performs Foxglove serialization via data_to_mcap.py.
VideoStage (video_stage.py): Stitches captured camera frames into .mp4 preview videos (dash + third-person).

`src/processing/physics.py` (The Math Layer)

Centralized source of truth for:
- Radial Velocity Injection: Projecting relative velocity onto LOS via calculate_radial_velocity().
- ADAS Metrics: Real-time Range, Azimuth, and Closing Velocity via calculate_relative_metrics().
- Actor Categorization: Classifying CARLA actors into vehicle, vru, pedestrian via get_actor_class().

`src/sensors.py` (SensorManager)

Manages Camera (RGB), Radar, and LiDAR.
Uses synchronized queues to ensure frame alignment.
Coordinate System: Left-handed (CARLA Default).

`scenarios/base.py` (The Plugin Contract)

All scenarios must inherit from ScenarioBase.
Required methods: setup(), step(), cleanup(), and name property.
Optional hooks: on_ego_spawned(), get_scenario_metadata(), max_frames, weather, ego_spawn_point.
Support for CLI parameter injection via apply_parameters(args.params).

🛰️ 3. Data & Visualization

Data Recording (`src/recorder.py`)

Saves images as .png, radar/lidar as .npy, and metadata as .jsonl.
Relative ADAS Metrics: Calculates range, azimuth, and closing velocity for all actors via src/processing/physics.py.

MCAP Conversion (`scripts/data_to_mcap.py`)

Converts raw dataset folders into Foxglove-compatible .mcap files.
Coordinate Conversion: Negates Y and Yaw to switch from CARLA left-handed to ROS/Foxglove right-handed convention.
Radar Math: Converts spherical (Depth/Az/Alt) to Cartesian (XYZ) for PointCloud representation.
Metrology Suite (v1.1):
- RD Heatmaps: dB-converted via 10*log10(data) - 68.0 system gain offset.
- RA Dual Plots: Static (absolute bounds) + Dynamic (auto-scaled peak tracking) via FastHeatmapEngine.
- CFAR Visualization: dB-converted threshold mask.
- Telemetry: Flattened metrics schema for native Foxglove Plot panel graphing.
- 3D FOV Frustums: Hardware-accurate LINE_LIST wireframes per radar type.

Visualization Engine (`scripts/ISOLATE/sim_radar_utils/plots.py`)

render_heatmap(): Stateless single-shot colormap rendering.
FastHeatmapEngine: Stateful matplotlib renderer that reuses figure memory for high-speed frame-by-frame rendering.
postprocess_ra(): Clutter subtraction, log compression, and fixed dynamic range clipping.
scan_convert_ra(): Polar-to-Cartesian sector projection for RA BEV display.

🖥️ 4. GUI Dashboard Architecture

Backend: Flask (app.py) running on port 5000.
Version: Orchestrator v1.1.
Simulator Lifecycle: Dashboard polls /api/simulator/status to detect if CARLA is Offline/Ready. Provides one-click initialization via the UI.
GPU Resource Control: Incorporates "Idle Mode" (Synchronous Throttling) to drop CARLA GPU usage to ~0% when inactive or while Shenron processing is active.
Streaming: Uses Server-Sent Events (SSE) to stream stdout to the browser.
Visual Feedback Tags: [SHENRON_INIT], [SHENRON_STEP], [AUTO-MCAP], [PROCESS_COMPLETED] drive progress bars and telemetry HUD.
Execution: Spawns run.bat as a subprocess.
Crucial Rule: Always set PYTHONUNBUFFERED=1 to prevent log delays in the GUI console.

🧠 5. Scenario Design & Best Practices

The "Deterministic Pattern"

Use manual coordinate-based control and set_target_velocity instead of Traffic Manager for repeatable maneuvers.
Spawn-and-Move: Spawn ego at a safe point, then set_transform to the exact start location.
Z-Axis Lift: Always spawn NPCs with a +0.5m offset to avoid ground-mesh collision errors ("Spot occupied").

Common Pitfalls

Stale Physics: Always world.tick() once after spawning the Ego but before starting scenario logic.
Graceful Shutdown: Use the "Stop Simulation" button in the dashboard. It writes a tmp/stop.flag which the orchestrator checks to ensure all recorders and data converters (Shenron/MCAP) finalize correctly.
Progress Bar Clash: Use pbar.write() for logs if a tqdm progress bar is active.
Asynchronous I/O: Use ThreadPoolExecutor for disk writes (images) to prevent simulation frame drops.

🚀 6. Execution Commands

Run Scenario: cmd /c run.bat <scenario_name> --frames <N> --params "KEY=VAL" --weather <Preset>
List Scenarios: cmd /c run.bat --list-scenarios
Run Testbench: cmd /c "call C:\ProgramData\miniconda3\Scripts\activate.bat carla312 && python scripts/test_shenron.py --iter <Name>"
Re-process MCAP Only: cmd /c run.bat --only-mcap --session data/<session_folder>

🤖 7. Agent Execution Rules

Shell Environment: This repository is hosted on Windows. All commands MUST use PowerShell syntax for general file operations and directory management.
Python Execution: All simulation and scenario commands MUST be launched via the run.bat orchestrator using the cmd /c prefix.
Standalone Scripts: Scripts like test_shenron.py or diagnostic tools must also use cmd /c with explicit environment activation to ensure compatibility:
- Correct: cmd /c "call C:\ProgramData\miniconda3\Scripts\activate.bat carla312 && python scripts/test_shenron.py --iter test"
- Incorrect: python scripts/test_shenron.py --iter test
Cmdlet Preference: Prioritize native PowerShell cmdlets or robust Windows utilities:
- Use Select-String or ripgrep instead of grep.
- Use ; to chain commands instead of &&.
- Use move/Copy-Item instead of mv/cp.
- Use tree /f for recursive directory visualization.

Generated by Antigravity AI | Last Updated: 2026-04-23 | Refer to .cursorrules for start-of-turn instructions.

9.6 KiB Raw Blame History