Simulation Meets Reality: Showcasing a Driver-in-the-Loop Simulation Platform for Autonomous Vehicles
In this demo, we demonstrate the closed-loop functionalities of the AVP, showcasing a Driver-in-the-loop configuration and perception model and retrieving sensor data in real time.
The Autonomous Vehicle Platform (AVP) is a complete hardware/software, X-in-the-loop capable platform (where X can be Model, Driver, Software, Hardware) allowing researchers and developers to design, train, and evaluate the performance of autonomous systems in a virtual environment. The platform includes several pre-configured and detailed 3D models of a sample city, vehicle dynamics, and sensor models and provides Perception, Fusion, and Driving functions as well. With complete control over weather conditions and traffic densities, developers can test autonomous vehicle behavior in various situations. Users can also collect raw data from the provided sensors, which include RGB camera, segmentation camera, LiDAR, radar, IMU, and GNSS, and uses them to further develop perception, fusion, and driving models. Finally, data on the vehicle's performance, such as speed, braking, and steering, can also be retrieved to identify improvement areas and ensure that the vehicle's behavior meets safety and performance standards.
