Leveraging LiDAR-Based Simulations to Quantify the Complexity of The Static Environment for Autonomous Vehicles in Rural Settings

• Author / Creator

  Abohassan, Mohamed

• The advent of autonomous vehicles promises to address key challenges faced by the transportation and traffic safety sectors, particularly concerning the high incidence of traffic fatalities and injuries. The full-scale deployment of autonomous vehicle fleets is contingent upon public acceptance, which is concerned about the available models' safety levels, evidenced by the multiple reported AV-related traffic incidents. Achieving the necessary safety standards for autonomous vehicles necessitates comprehensive and rigorous testing. Virtual simulations take precedence over road and track testing, as they expedite AV development by enabling safe, cost-effective, and large-scale testing.
  AVs face a big data problem stemming from the vast volume of collected sensory information data surpassing real time processing capabilities. This mismatch often leads to driving errors and traffic collisions. Although different models to quantify the complexity of the environment have been explored, they often lacked objectivity, ignored the physics of the simulated sensors, and failed to utilize robust LiDAR data. Hence, this study aims to develop a framework that can transform the complexity of the environment into easy-to-understand values while accounting for the drawbacks of the existing frameworks.
  This study conducted virtual simulations on 34 kilometers of LiDAR-based digital environments sourced from two-way-two-lane rural roads in Alberta, Canada. The complexity of the environment was assessed by calculating the required real time data rates essential for AVs to maintain their regular navigation functions. The data rate requirements encompass the sensor's specifications, the surrounding environment's dynamicity, and weather conditions in its calculations. Furthermore, this study focused on the static physical environment only, dividing the analyzed environments into RRFs and FE to isolate and study the impact of general road geometry features such as vertical curves, horizontal curves, and roadway width on the complexity of the environment and quantify the extra computational burdens incurred from the analysis of the entire section to highlight the problem of WVC for AVs.
  This study employed two distinct approaches to demonstrate the versatility of the implemented framework and explore diverse perspectives on environment perception for AVs: the primary occupancy method and the secondary volumetric method, where the advantages and drawbacks of each method are highlighted.
  The study analyzed fluctuations in data rate requirements along the vehicle's trajectory by dividing the surveyed roadways into distinct frames and generating novel synthetic viewpoints to faithfully replicate real-world situations. This process was facilitated using the open-source Vista simulator. The analysis revealed that the environment could be deemed complex in two scenarios. Firstly, when data rate values spiked, it indicated a substantial volume of information that needed processing. Secondly, when data rates plummeted, it signified a significant loss of information, potentially jeopardizing the vehicle's operations.
  The analysis results indicated that the roadside features are anticipated to escalate environmental complexity by 140-400% based on density. Additionally, widening the roadway by adding an extra lane was observed to raise processing requirements by 12.3-16.5%. Concerning road alignment, crest vertical curves were found to decrease data rates by up to 4% due to occlusion challenges at these points, while sag curves increased requirements by 7% due to enhanced visibility. In horizontal curves, roadside occlusion contributed to a decrease in data rate requirements by up to 19%. As for the weather conditions, heavy rain increased the AV's processing demands by a significant 240% when compared to normal weather conditions.
  The developed framework and results, supported by statistical testing, can help AV developers make more informed decisions by understanding the impact of the different road elements. Moreover, government agencies and IOOs can also exploit the findings of this study to accommodate AV requirements in the current human-tailored road designs and optimize future designs for AV deployment.

• Subjects / Keywords

  • Transportation Engineering
  • Autonomous Vehciles
  • Digital Twins

• Graduation date

  Spring 2024

• Type of Item

  Thesis

• Degree

  Master of Science

• DOI

  https://doi.org/10.7939/r3-4551-7729

• License

  This thesis is made available by the University of Alberta Libraries with permission of the copyright owner solely for non-commercial purposes. This thesis, or any portion thereof, may not otherwise be copied or reproduced without the written consent of the copyright owner, except to the extent permitted by Canadian copyright law.