Overview
This project implements an autonomous parallel parking system in the CARLA Simulator using a rule-based control approach.
The system detects parking slots using LiDAR data and executes a complete parking maneuver using a Finite State Machine (FSM) and handcrafted control strategies.
Unlike learning-based methods, this approach is:
- Lightweight
- Interpretable
- Real-time capable
Key Features
Autonomous parallel parking LiDAR-based slot detection Finite State Machine (FSM) control Safety-aware motion using obstacle detection No ROS required (pure Python implementation)
System Architecture
The system consists of the following components:
-
Slot Detector
- Processes LiDAR point cloud
- Identifies valid parking gaps
-
Finite State Machine (FSM)
- Controls the parking maneuver:
- Scanning
- Forward alignment
- Reverse (Phase 1 & 2)
- Final centering
- Controls the parking maneuver:
-
Safety Monitor
- Prevents collisions using LiDAR-based clearance checks
-
Controller
- Speed PID + steering logic
SCANNING → FORWARD ALIGN → STOP → REVERSE (Phase 1) → REVERSE (Phase 2) → CENTERING → PARKED
Requirements
- Python 3.8+
- CARLA 0.9.15
- NumPy
- OpenCV
How to Run
- Start CARLA: CarlaUE4.exe
- Open Anaconda Prompt
- go to the location of your .py file
- activate carla-sim
- Run the script: Autonomous_Parking_CARLA.py
Key Parameters
| Parameter | Description |
|---|---|
MIN_SLOT_LENGTH |
Minimum gap for parking |
TARGET_EX_TOL |
Longitudinal tolerance |
TARGET_EY_TOL |
Lateral tolerance |
TARGET_YAW_TOL |
Orientation tolerance |
Results
The system successfully:
- Detects parking slots using real-time LiDAR
- Performs smooth multi-stage parking
- Avoids collisions using safety constraints
Methodology
This project uses a rule-based approach combining:
- Geometric reasoning
- Sensor-based perception
- Finite State Machine control
This avoids:
- Long training times
- High computational cost of deep learning
Future Improvements
- Improve trajectory optimization (MPC)
- Extend to perpendicular parking
- Integrate learning-based perception
#Author Salma Diaa PhD Student – Smart Cities Western University