Collaborative heterogeneous 3D scene representations for Urban Mobility

  The C3DRUM project focuses on urban mobility, particularly micromobility, using computer vision and artificial intelligence methods. Urban mobility significantly impacts the quality of life for city residents, environmental protection, and the economic growth of communities. Achieving effective urban mobility requires promoting transportation methods and policies that are safe, sustainable, efficient, competitive, and inclusive.

   Mobility in large cities presents significant challenges, including environmental pollution and increased commuting times. To address these issues, cities such as Barcelona are increasingly incorporating solutions like enhancing public transportation and promoting micromobility (e.g., emission-free vehicles such as e-bikes and e-scooters) into their Sustainable Urban Mobility Plans (SUMPs). While these measures help reduce carbon emissions, they also introduce new challenges. For example, micromobility options are often perceived as unsafe and can create conflicts with pedestrians or motor vehicles, while public transportation systems often lack the flexibility to adapt to dynamic passenger flows and may be slower than other modes, reducing their attractiveness to users.

   Artificial intelligence (AI) and computer vision (CV) have the potential to transform urban mobility by addressing these challenges, such as improving safety, dynamically optimizing transportation flows, and enabling better space allocation for parking. However, current AI and CV methods are not yet advanced enough to handle the full complexity of urban scenarios, especially in dynamic environments.

   Our goal is to explore innovative algorithms that can lead to more efficient, sustainable, and safer transportation solutions. In this context, CV holds significant potential due to its ability to perceive the environment, capture critical data, and analyze it to inform decision-making and improve mobility systems. Although many efforts have been made in this field, the problems are very complex, and in many cases, the available algorithms and tools are not powerful enough to overcome this complexity. New 3D scene representations based on collaborative multimodal fusion from heterogeneous sensors have the potential to greatly enhance analysis, but they are not yet mature enough for real-world applications.

   C3DRUM aims to create new or improve existing 3D scene representations, as well as to propose novel algorithms and deep neural networks for analyzing them. With these new representations and algorithms, we will develop more powerful urban mobility applications. For example, determining the density of people in the path of micro-vehicles, analyzing the behavior of people at bus stops, guiding vehicles in urban scenes, or analyzing defects in bike lanes and traffic signs.