Skip to main content
eScholarship
Open Access Publications from the University of California

Scholarly Works (4 results)

  • Thesis
  • Peer Reviewed
UC San Diego Electronic Theses and Dissertations (2025)

Autonomous driving technologies have evolved rapidly recently. However, a key bottleneck is the reliance on high-definition (HD) maps. These maps provide abundant features such as lane lines and lane connectivities with centimeter-level accuracy. They are crucial to downstream tasks such as localizing the ego vehicle, predicting agent trajectories, and path planning. Nevertheless, it is costly to create and maintain HD maps, often involving time-consuming data collection and labor-intensive labeling. This presents a challenge for autonomous driving to scale up rapidly.

We investigate automating the offline map generation or creating maps online to replace HD maps. To generate robust offline maps, we use a probabilistic semantic grid to accumulate semantic labels from point clouds fused with semantic images. We incorporated the confusion matrix for the semantic segmentation model which significantly improved accuracy for the challenging lane boundary class.

While offline maps remove human annotation, it still require data collection to maintain up-to-date maps. Online mapping, which builds maps while driving, addresses this problem. Recent advances in low-cost vision-only online mapping increase the adoption. However, the maps built online have limited range due to heavy occlusion. We proposed using low-maintenance standard-definition (SD) maps, such as OpenStreetMaps (OSM), to enhance online mapping. We show that integrating SD maps via rasterized approach provides performance boosts and the graph-based approach can additionally yield a more lightweight network without sacrificing the performance.

As online mapping gains popularity, challenges arise when deploying the algorithm on a wide range of platforms. The performance of the model degrades significantly when the camera parameters and locations change. It is expensive to collect and annotate the data to retrain a new model for each set of sensor configurations. We propose to use sensor configuration agnostic semantic mapping to generate an intermediate semantic map to bridge the representation gap and show that it increases the zero-shot performance transfer significantly. Additionally, we created a pipeline to use Gaussian splitting to synthesize novel view data to recreate the dataset from a different sensor configuration. We show that this new dataset is effective for augmenting the data or pretraining.

    Cover page: Towards Robust and Scalable Mapping for Autonomous Driving
    • Article
    • Peer Reviewed
    Proceedings of the Annual Meeting of the Cognitive Science Society, Volume 46 (2024)
    The multimodal deep neural networks, represented by CLIP, have generated rich downstream applications owing to their excellent performance, thus making understanding the decision-making process of CLIP an essential research topic. Due to the complex structure and the massive pre-training data, it is often regarded as a black-box model that is too difficult to understand and interpret. Concept-based models map the black-box visual representations extracted by deep neural networks onto a set of human-understandable concepts and use the concepts to make predictions, enhancing the transparency of the decision-making process. However, these methods involve the datasets labeled with fine-grained attributes by expert knowledge, which incur high costs and introduce excessive human prior knowledge and bias. In this paper, we observe the long-tail distribution of concepts, based on which we propose a two-stage Concept Selection Model (CSM) to mine core concepts without introducing any human priors. The concept greedy rough selection algorithm is applied to extract head concepts, and then the concept mask fine selection method performs the extraction of core concepts. Experiments show that our approach achieves comparable performance to end-to-end black-box models, and human evaluation demonstrates that the concepts discovered by our method are interpretable and comprehensible for humans.
      Cover page: Understanding Multimodal Deep Neural Networks: A Concept Selection ViewCreative Commons 'BY' version 4.0 license
      • Article
      • Peer Reviewed
      UC San Diego Previously Published Works (2023)
      Statistical learning techniques and increased computational power have facilitated the development of self-driving car technology. However, a limiting factor has been the high expense of scaling and maintaining high-definition (HD) maps. These maps are a crucial backbone for many approaches to self-driving technology. In response to this challenge, we present an approach that fuses pre-built point cloud map data with images to automatically and accurately identify static landmarks such as roads, sidewalks, and crosswalks. Our pipeline utilizes semantic segmentation of 2D images, associates semantic labels with points in point cloud maps to pinpoint locations in the physical world, and employs a confusion matrix formulation to generate a probabilistic bird's-eye view semantic map from semantic point clouds. The approach has been tested in an urban area with different segmentation networks to generate a semantic map with road features. The resulting map provides a rich context of the environment that is valuable for downstream tasks such as trajectory generation and intent prediction. Moreover, it has the potential to be extended to the automatic generation of HD maps for semantic features. The entire software pipeline is implemented in the robot operating system (ROS), a widely used robotics framework, and made available.
        Cover page: Probabilistic Semantic Mapping for Autonomous Driving in Urban EnvironmentsCreative Commons 'BY' version 4.0 license
        • Article
        • Peer Reviewed
        UC San Diego Previously Published Works (2023)
        This work explores methodologies for dynamic trajectory generation for urban driving environments by utilizing coarse global plan representations. In contrast to state-of-the-art architectures for autonomous driving that often leverage lane-level high-definition (HD) maps, we focus on minimizing required map priors that are needed to navigate in dynamic environments that may change over time. To incorporate high-level instructions (i.e., turn right vs. turn left at intersections), we compare various representations provided by lightweight and open-source OpenStreetMaps (OSM) and formulate a conditional generative model strategy to explicitly capture the multimodal characteristics of urban driving. To evaluate the performance of the models introduced, a data collection phase is performed using multiple full-scale vehicles with ground truth labels. Our results show potential use cases in dynamic urban driving scenarios with real-time constraints. The dataset is released publicly as part of this work in combination with code and benchmarks.
          Cover page: Conditional Generative Models for Dynamic Trajectory Generation and Urban DrivingCreative Commons 'BY' version 4.0 license
          Top