nfiniity

In an era of rapid urbanization and technological advancement, the challenge of creating efficient, safe, and inclusive public transportation systems has never been more pressing. At the Chair of Traffic Engineering and Control at Technical University of Munich (TUM), we've been exploring innovative solutions to address this challenge, focusing on the integration of Collective Perception Systems (CPS) in automated public transport. This work has been executed within the scope of the research project ‘OBACHT’, which is funded by the TUM Georg Nemetschek Institute ‘Artificial Intelligence for the Built World’. Our research aims to enhance the capabilities of automated vehicles, particularly in detecting and responding to vulnerable road users in complex urban environments.

The Challenge

Traditional automated vehicles often struggle with occlusion and limited sensor range, especially in bustling urban settings. This can lead to safety concerns, particularly for mobility-impaired users who may not be easily detected by vehicle-based sensors alone. Our goal was to develop a system that could overcome these limitations, improving both the safety and inclusivity of automated public transport.

Our Approach

Public Transport Infrastructure Control System (PT-ICS): To address these challenges, we developed a PT-ICS that integrates advanced sensing technologies, Vehicle-to-Everything (V2X) communications, and data fusion strategies. Central to our V2X communication setup is the cube:evk device from cubesys GmbH, a subsidiary from nfiniity GmbH, which enabled real-time information exchange between infrastructure, vehicles, and vulnerable road users (VRUs).

The Role of cube:evk

The cube:evk served as our primary V2X communication module, facilitating the transmission of critical messages such as Collective Perception Messages (CPMs), VRU Awareness Messages (VAMs), and Cooperative Awareness Messages (CAMs). Key features that benefited our research include:

1. Dual-mode operation: Supporting both ITS-G5 and C-V2X standards
2. Compatibility with open-source V2X stacks (Vanetza)
3. Integration with ROS 2 and the Autoware autonomous driving stack

Experimental Setup and Results

We developed three proxy systems to test our PT-ICS:

Mobile LiDAR Station (Infrastructure Proxy)

• Purpose: To represent roadside infrastructure capable of detecting and communicating with vehicles and VRUs.
• Key Components: Horizon Gen2 LiDAR by Ouster • Robosense bpearl Super Wide FOV Short-range Blind Spot LiDAR • ZED2i stereo camera from Stereolabs • GNSS integration with H-RTK F9P GNSS Series module • cube:evk for V2X communication
• Capabilities: 3D object detection, high-precision localization, and real-time data transmission.

Automated Rickshaw (Public Transport Vehicle Proxy)

• Purpose: To simulate an automated public transport vehicle.
• Key Components: • OS1-32-Uniform Gen2 Rev6 Scanning LiDAR • ZED2i stereo camera • Xsens MTi-680G RTK GNSS/INS for high-accuracy localization • cube:evk for V2X communication
• Capabilities: Autonomous navigation, object detection, and V2X message processing.

V2X-equipped Scooter (Connected VRU Proxy)

• Purpose: To represent a connected vulnerable road user.
• Key Components: • Embedded computer with display unit • u-blox ZED-F9R GNSS module for precise positioning • Unex OBU-352 for V2X communication
• Capabilities: Real-time location tracking and V2X message transmission.

The overall system architecture:

Our experiments focused on two key scenarios:

Detection of Connected VRUs

• Setup: The V2X-equipped scooter was positioned in an area occluded from the direct view of the automated rickshaw.
• Process: The scooter continuously broadcast its position via VAMs.
• Objective: To assess the system's ability to detect and track a connected VRU beyond the vehicle's direct line of sight.
• Key Measurements: Accuracy of position reporting, latency of communication, and successful integration of VRU data into the vehicle's perception system.

Detection of Mobility-Impaired Users Through Infrastructure

• Setup: A pedestrian (simulating a mobility-impaired user) was positioned in an area visible to the mobile LiDAR station but occluded from the automated rickshaw.
• Process: The mobile LiDAR station detected the pedestrian, created a CPM, and transmitted this information to the automated rickshaw.
• Objective: To evaluate the effectiveness of infrastructure-based sensing in expanding the vehicle's perception capabilities.
• Key Measurements: Detection accuracy of the infrastructure sensors, successful transmission of CPMs, and the vehicle's ability to incorporate this external data into its environment model.

Key Results

1. Connected VRU Detection:
   • The system successfully detected and tracked the V2X-equipped scooter in real-time.
   • VAMs transmitted by the scooter were accurately received and processed by the infrastructure and the automated vehicle.
   • The use of higher-precision GNSS modules significantly improved localization accuracy, enhancing the system's overall performance.
2. Infrastructure-based Detection of Mobility-Impaired Users:
   • The mobile LiDAR station successfully detected occluded pedestrians (representing mobility-impaired users) that were not visible to the automated vehicle.
   • This information was effectively communicated to the automated vehicle via CPMs, enhancing its perception capabilities.
   • The system demonstrated the ability to extend the vehicle's field of view beyond its direct line of sight, crucial for safer urban navigation.
3. V2X Communication Performance:
   • The cube:evk devices consistently facilitated low-latency communication between all components of the system, including interoperability with other V2X systems.
   • CPMs and VAMs were reliably transmitted and received, enabling real-time data exchange crucial for collective perception.
4. Integration with Autoware:
   • The V2X data was successfully integrated into the Autoware stack, demonstrating the potential for seamless incorporation of collective perception into existing autonomous driving platforms.

These results showcase the potential of CPS in enhancing the perception capabilities of automated public transport systems, particularly in complex urban environments where traditional sensing methods may fall short.

Future Directions

Building on these promising results, future work will focus on:

1. Expanding the range of detectable VRU types, with a particular emphasis on mobility-impaired users.
2. Enhancing data fusion algorithms to better integrate information from diverse sources.
3. Conducting larger-scale trials in varied urban environments to further validate the system's performance.

Conclusion

Our research demonstrates the significant potential of integrating CPS in automated public transport systems. The cube:evk devices were instrumental in realizing this integration, providing reliable and flexible V2X communication capabilities. As V2X technologies continue to evolve, we anticipate further improvements in urban mobility safety and inclusivity.

About

The Chair of Traffic Engineering and Control at TUM is engaged in research and teaching on methods and technologies for the collection of traffic data as well as the description and spatio-temporal management and control of traffic, both for private and public transportation.

Mario Ilic
TUM Chair of Traffic Engineering and Control
mario.ilic@tum.de