Radar-Triggered Video Recording System Using CARLA Simulator
Background
This project aimed to implement a system where radar sensors detect objects and trigger a TI board to record video for 10 seconds. Developed as part of an Australian initiative, the solution was designed and validated in the CARLA simulator, offering a cost-effective platform for testing radar and camera integration under diverse conditions.
Problem Statement
The objective was to create a system where radar sensors could detect objects within a specific range and trigger a video recording on a TI board. Key challenges included:
Ensuring accurate object detection in various environmental conditions.
Establishing reliable communication between the radar and the TI board to initiate video recording.
Testing the system in diverse scenarios without requiring physical hardware.
Implementation
Simulator Setup:
The CARLA simulator was used to replicate real-world scenarios.
A fixed location on the Town10 map was chosen to mimic a radar mounted on a building.
Radar Configuration:
A radar sensor with a 45° horizontal field of view and a 25-meter range was set up in the simulation.
The radar was programmed to detect objects and compute their position based on azimuth, altitude, and depth values.
Detected objects within 20 meters were flagged for triggering the TI board.
TI Board Communication:
The radar sensor sent a trigger signal (object_detected) to the TI board using a socket-based communication protocol.
Upon receiving the signal, the TI board initiated a 10-second video recording from its connected camera.
Testing in Diverse Conditions:
The system was tested under various weather conditions, including rain, fog, and clear skies, as well as different times of the day such as morning, noon, evening, and night.
These scenarios helped ensure consistent performance and reliability across changing environmental factors.
Validation:
Debugging tools in CARLA visualized radar-detected points to verify correct detections.
Communication logs confirmed successful trigger signals and timely video recording initiation.
Results
The radar sensor consistently detected objects within the defined range, even under adverse weather conditions and varying lighting. The TI board reliably received trigger signals and initiated recordings as expected. The system was tested across multiple scenarios in CARLA, with environmental variables like weather and time of day taken into account.
Below is an image from the CARLA simulator showing the radar detection area, represented by a defined range of 20 meters. This visual illustrates the detection coverage area within which the radar sensor identifies objects, ensuring reliable operation in varied conditions.
Conclusion
This case study highlights the successful development of a radar-triggered video recording system. By leveraging the CARLA simulator, the team ensured accurate object detection and seamless radar-to-TI board integration, with consistent performance tested under multiple weather conditions and times of day