PROSENSE: PROactive SENSing for autonomous driving

Purpose and goal

Despite rapid research & innovation in scene perception for autonomous driving, many challenges remain open in understanding complex traffic scenes with occluded road users. To overcome these challenges, Prosense aims to extend the state-of-the-art by: 1. Incorporating multi-sensor information to create a scene representation that includes possible occlusions. 2. Enhancing the robustness of scene perception in different traffic scenarios. 3. Integrating scene metadata for context-aware detection & classification of occluded road users, object anticipation & prediction.

Expected results and effects

At the end of the project, the advancements in key technology areas will extend the scene perception capabilities.The main results of this project are: 1. Developed algorithms for handling occlusions based on multi-modal sensor data and scene metadata. 2. Integration of advanced algorithms for scene perception in the presence of occluded objects on-board an autonomous research vehicle. 3. Pubic demonstration of integrated algorithms emulating complex traffic occlusions.

Planned approach and implementation

The project will start by focusing on the two main project enablers, generating synthetic data and developing a 3D occlusion model. Meanwhile, fundamental research on multi-source fusion for robust real-time scene perception and proactive context aware object perception will be conducted. This results in continuous development & assessment of algorithms. The outcome will be published in reputable conferences and journals. As new algorithms & functionalities are developed, they will be integrated in autonomous driving platform for timely demonstration on a public test track.