arXiv subjects

Jingyu Song

Publications and source records attributed to Jingyu Song.

13 recordsLinked to original sources

arXiv2026

CRISP: A Spatiotemporal Camera-Radar Backbone for Driving via Forecasting-Based World-Model Pretraining

Camera-radar (CR) fusion is a practical sensing configuration for autonomous driving, but existing models are typically trained with task-specific supervision, limiting reusable representation learning. We present CRISP, a spatiotemporal CR backbone pretrained through forecasting-based representation learning. Given historical multi-view images and radar sweeps, CRISP learns a unified bird's-eye-view (BEV) representation by predicting future LiDAR point clouds. LiDAR is used only as privileged supervision during pretraining; the deployed model requires only camera and radar. To make forecasting-based pretraining effective for CR fusion, CRISP introduces an enhanced radar encoder, radar-enhanced temporal self-attention, and multimodal feature rendering with modality innovation gating. These components inject radar range and Doppler cues into BEV temporal propagation and allow BEV tokens to selectively incorporate camera and radar evidence. Experiments on nuScenes show that CRISP improves long-horizon point cloud forecasting and transfers effectively to downstream tasks, including 3D detection, tracking, online mapping, motion forecasting, future occupancy prediction, and planning, suggesting that predictive CR pretraining is a promising path toward scalable driving representations under practical sensor configurations. The project website is https://umfieldrobotics.github.io/CRISP.

cs.CV↗

arXiv2025

CHyLL: Learning Continuous Neural Representations of Hybrid Systems

Learning the flows of hybrid systems that have both continuous and discrete time dynamics is challenging. The existing method learns the dynamics in each discrete mode, which suffers from the combination of mode switching and discontinuities in the flows. In this work, we propose CHyLL (Continuous Hybrid System Learning in Latent Space), which learns a continuous neural representation of a hybrid system without trajectory segmentation, event functions, or mode switching. The key insight of CHyLL is that the reset map glues the state space at the guard surface, reformulating the state space as a piecewise smooth quotient manifold where the flow becomes spatially continuous. Building upon these insights and the embedding theorems grounded in differential topology, CHyLL concurrently learns a singularity-free neural embedding in a higher-dimensional space and the continuous flow in it. We showcase that CHyLL can accurately predict the flow of hybrid systems with superior accuracy and identify the topological invariants of the hybrid systems. Finally, we apply CHyLL to the stochastic optimal control problem.

cs.LG↗

arXiv2025

FishDetector-R1: Unified MLLM-Based Framework with Reinforcement Fine-Tuning for Weakly Supervised Fish Detection, Segmentation, and Counting

Analyzing underwater fish imagery is critical for ecological monitoring but remains difficult due to visual degradation and costly annotations. We introduce FishDetector-R1, a unified MLLM-based framework for fish detection, segmentation, and counting under weak supervision. On the DeepFish dataset, our framework achieves substantial gains over baselines, improving AP by 20% and mIoU by 10%, while reducing MAE by 30% and GAME by 35%. These improvements stem from two key components: a novel detect-to-count prompt that enforces spatially consistent detections and counts, and Reinforcement Learning from Verifiable Reward (RLVR) with a complementary scalable paradigm leveraging sparse point labels. Ablation studies further validate the effectiveness of this reward design. Moreover, the improvement generalizes well to other underwater datasets, confirming strong cross-domain robustness. Overall, FishDetector-R1 provides a reliable and scalable solution for accurate marine visual understanding via weak supervision. The project page for FishDetector-R1 is https://umfieldrobotics.github.io/FishDetector-R1.

cs.CV↗

arXiv2025

Zero-Human Demonstration End-to-end Autonomous Driving with Trajectory Scorer

Human demonstrations are widely considered the cornerstone of end-to-end (E2E) autonomous driving despite human demonstration's scarcity for long-tail and safety-critical scenarios. Nonetheless, current E2E autonomous driving (AD) training paradigms continue to rely on human demonstrations. Imitation learning (IL) requires human demonstrations for training, whereas reinforcement learning (RL) has emerged as a promising alternative to reduce this dependency. However, most existing RL methods for E2E AD still rely implicitly on human demonstrations. A pure rewards-based RL method can overcome the need for human demonstrations, but general RL policy gradient methods suffer from the cold-start problem. In this paper, we propose ZTRS (Zero-human demonstration end-to-end autonomous driving with TRajectory Scorer) - a complete RL-based E2E planning paradigm trained solely on real-world images and rule-based rewards, entirely without human demonstration. Through our proposed Exhaustive Policy Optimization (EPO), a policy gradient variant tailored for enumerable trajectory actions and dense supervision, ZTRS enables the model to generalize better to long-tail driving scenarios. We demonstrate this generalization through our SOTA performance against IL approaches on both long-tail Navhard and closed-loop HUGSIM datasets. Project page: https://zhenxinli.net/ZTRS/.

cs.RO↗

arXiv2025

DriveCritic: Towards Context-Aware, Human-Aligned Evaluation for Autonomous Driving with Vision-Language Models

Benchmarking autonomous driving planners to align with human judgment remains a critical challenge, as state-of-the-art metrics like the Extended Predictive Driver Model Score (EPDMS) lack context awareness in nuanced scenarios. To address this, we introduce DriveCritic, a novel framework featuring two key contributions: the DriveCritic dataset, a curated collection of challenging scenarios where context is critical for correct judgment and annotated with pairwise human preferences, and the DriveCritic model, a Vision-Language Model (VLM) based evaluator. Fine-tuned using a two-stage supervised and reinforcement learning pipeline, the DriveCritic model learns to adjudicate between trajectory pairs by integrating visual and symbolic context. Experiments show DriveCritic significantly outperforms existing metrics and baselines in matching human preferences and demonstrates strong context awareness. Overall, our work provides a more reliable, human-aligned foundation to evaluating autonomous driving systems. The project page for DriveCritic is https://song-jingyu.github.io/DriveCritic

cs.CV↗

arXiv2025

OceanSim: A GPU-Accelerated Underwater Robot Perception Simulation Framework

Underwater simulators offer support for building robust underwater perception solutions. Significant work has recently been done to develop new simulators and to advance the performance of existing underwater simulators. Still, there remains room for improvement on physics-based underwater sensor modeling and rendering efficiency. In this paper, we propose OceanSim, a high-fidelity GPU-accelerated underwater simulator to address this research gap. We propose advanced physics-based rendering techniques to reduce the sim-to-real gap for underwater image simulation. We develop OceanSim to fully leverage the computing advantages of GPUs and achieve real-time imaging sonar rendering and fast synthetic data generation. We evaluate the capabilities and realism of OceanSim using real-world data to provide qualitative and quantitative results. The code and detailed documentation are made available on the project website to support the marine robotics community: https://umfieldrobotics.github.io/OceanSim.

cs.RO↗

arXiv2024

MemFusionMap: Working Memory Fusion for Online Vectorized HD Map Construction

High-definition (HD) maps provide environmental information for autonomous driving systems and are essential for safe planning. While existing methods with single-frame input achieve impressive performance for online vectorized HD map construction, they still struggle with complex scenarios and occlusions. We propose MemFusionMap, a novel temporal fusion model with enhanced temporal reasoning capabilities for online HD map construction. Specifically, we contribute a working memory fusion module that improves the model's memory capacity to reason across a history of frames. We also design a novel temporal overlap heatmap to explicitly inform the model about the temporal overlap information and vehicle trajectory in the Bird's Eye View space. By integrating these two designs, MemFusionMap significantly outperforms existing methods while also maintaining a versatile design for scalability. We conduct extensive evaluation on open-source benchmarks and demonstrate a maximum improvement of 5.4% in mAP over state-of-the-art methods. The project page for MemFusionMap is https://song-jingyu.github.io/MemFusionMap

cs.CV↗

arXiv2024

TURTLMap: Real-time Localization and Dense Mapping of Low-texture Underwater Environments with a Low-cost Unmanned Underwater Vehicle

Significant work has been done on advancing localization and mapping in underwater environments. Still, state-of-the-art methods are challenged by low-texture environments, which is common for underwater settings. This makes it difficult to use existing methods in diverse, real-world scenes. In this paper, we present TURTLMap, a novel solution that focuses on textureless underwater environments through a real-time localization and mapping method. We show that this method is low-cost, and capable of tracking the robot accurately, while constructing a dense map of a low-textured environment in real-time. We evaluate the proposed method using real-world data collected in an indoor water tank with a motion capture system and ground truth reference map. Qualitative and quantitative results validate the proposed system achieves accurate and robust localization and precise dense mapping, even when subject to wave conditions. The project page for TURTLMap is https://umfieldrobotics.github.io/TURTLMap.

cs.RO↗

arXiv2024

CRKD: Enhanced Camera-Radar Object Detection with Cross-modality Knowledge Distillation

In the field of 3D object detection for autonomous driving, LiDAR-Camera (LC) fusion is the top-performing sensor configuration. Still, LiDAR is relatively high cost, which hinders adoption of this technology for consumer automobiles. Alternatively, camera and radar are commonly deployed on vehicles already on the road today, but performance of Camera-Radar (CR) fusion falls behind LC fusion. In this work, we propose Camera-Radar Knowledge Distillation (CRKD) to bridge the performance gap between LC and CR detectors with a novel cross-modality KD framework. We use the Bird's-Eye-View (BEV) representation as the shared feature space to enable effective knowledge distillation. To accommodate the unique cross-modality KD path, we propose four distillation losses to help the student learn crucial features from the teacher model. We present extensive evaluations on the nuScenes dataset to demonstrate the effectiveness of the proposed CRKD framework. The project page for CRKD is https://song-jingyu.github.io/CRKD.

cs.CV↗

arXiv2024

LiRaFusion: Deep Adaptive LiDAR-Radar Fusion for 3D Object Detection

We propose LiRaFusion to tackle LiDAR-radar fusion for 3D object detection to fill the performance gap of existing LiDAR-radar detectors. To improve the feature extraction capabilities from these two modalities, we design an early fusion module for joint voxel feature encoding, and a middle fusion module to adaptively fuse feature maps via a gated network. We perform extensive evaluation on nuScenes to demonstrate that LiRaFusion leverages the complementary information of LiDAR and radar effectively and achieves notable improvement over existing methods.

cs.RO↗

arXiv2023

Uncertainty-Aware Acoustic Localization and Mapping for Underwater Robots

For underwater vehicles, robotic applications have the added difficulty of operating in highly unstructured and dynamic environments. Environmental effects impact not only the dynamics and controls of the robot but also the perception and sensing modalities. Acoustic sensors, which inherently use mechanically vibrated signals for measuring range or velocity, are particularly prone to the effects that such dynamic environments induce. This paper presents an uncertainty-aware localization and mapping framework that accounts for induced disturbances in acoustic sensing modalities for underwater robots operating near the surface in dynamic wave conditions. For the state estimation task, the uncertainty is accounted for as the added noise caused by the environmental disturbance. The mapping method uses an adaptive kernel-based method to propagate measurement and pose uncertainty into an occupancy map. Experiments are carried out in a wave tank environment to perform qualitative and quantitative evaluations of the proposed method. More details about this project can be found at https://umfieldrobotics.github.io/PUMA.github.io.

cs.RO↗

arXiv2023

Convolutional Bayesian Kernel Inference for 3D Semantic Mapping

Robotic perception is currently at a cross-roads between modern methods, which operate in an efficient latent space, and classical methods, which are mathematically founded and provide interpretable, trustworthy results. In this paper, we introduce a Convolutional Bayesian Kernel Inference (ConvBKI) layer which learns to perform explicit Bayesian inference within a depthwise separable convolution layer to maximize efficency while maintaining reliability simultaneously. We apply our layer to the task of real-time 3D semantic mapping, where we learn semantic-geometric probability distributions for LiDAR sensor information and incorporate semantic predictions into a global map. We evaluate our network against state-of-the-art semantic mapping algorithms on the KITTI data set, demonstrating improved latency with comparable semantic label inference results.

cs.RO↗

arXiv2022

MotionSC: Data Set and Network for Real-Time Semantic Mapping in Dynamic Environments

This work addresses a gap in semantic scene completion (SSC) data by creating a novel outdoor data set with accurate and complete dynamic scenes. Our data set is formed from randomly sampled views of the world at each time step, which supervises generalizability to complete scenes without occlusions or traces. We create SSC baselines from state-of-the-art open source networks and construct a benchmark real-time dense local semantic mapping algorithm, MotionSC, by leveraging recent 3D deep learning architectures to enhance SSC with temporal information. Our network shows that the proposed data set can quantify and supervise accurate scene completion in the presence of dynamic objects, which can lead to the development of improved dynamic mapping algorithms. All software is available at https://github.com/UMich-CURLY/3DMapping.

cs.CV↗