2024 AutoTest Technical Conference

Modeling and verification of lidar sensor data and its internal algorithmic modules are crucial for robust product development. Prior studies have Explored lidar sensor modeling, Often lacking noise characterization. In addition, phenomenological modeling has been explored for various lidar applications. However, there exists a gap concerning the fusion of point cloud generation and phenomenological modeling, alongside their comprehensive verification against actual sensor data. The paper in hand proposes two key models: a generative AI model for lidar point cloud generation and a phenomenological model for algorithmic modeling such as object detection, object tracking, and lane detection. These models are rigorously verified and validated against real sensor data. By integrating these models, our approach enables both end-to-end testing for algorithmic processes and targeted testing of internal modules. This flexibility facilitates for thorough validation of perception algorithms and promotes efficient development. Verification results show a high level of similarity between our models and actual sensor data. The phenomenological model achieves an average similarity of 95% with the real sensor, showcasing the fidelity of our developed model. Similarly, the generative AI model accurately captures the spatial and intensity characteristics and noise profile of lidar point clouds, with an average similarity of 96%. The fusion ensures model robustness and reliability. (Ragaby, Basem and Adham) This paper proposes a method for rigorously validating LIDAR systems using simulation models. By simulating real-world scenarios, including environmental conditions and sensor characteristics, we can thoroughly test the performance of LIDAR systems. Through systematic experimentation and analysis, we identify potential limitations and optimize sensor configurations to improve overall system reliability. This approach offers a cost-effective and scalable solution, ensuring the development of dependable LIDAR systems across various applications. By leveraging simulation models, we can streamline the validation process and accelerate the deployment of LIDAR technology in diverse industries.