The field of autonomous perception and localization is rapidly advancing, with a focus on developing more accurate and efficient methods for environmental mapping, object detection, and scene understanding. Recent research has explored the use of camera-only systems, which can reduce costs and improve flexibility, as well as the integration of various sensors such as LiDAR, radar, and inertial measurement units. Notable papers in this area include 'Camera-Only Bird's Eye View Perception: A Neural Approach to LiDAR-Free Environmental Mapping for Autonomous Vehicles', which proposes a camera-only perception framework for autonomous vehicles, and 'RESAR-BEV: An Explainable Progressive Residual Autoregressive Approach for Camera-Radar Fusion in BEV Segmentation', which presents a progressive refinement framework for camera-radar fusion in BEV segmentation. Other promising approaches involve the use of transformer-based architectures, such as 'Transformer-Based Dual-Optical Attention Fusion Crowd Head Point Counting and Localization Network', and the development of more robust and generalizable methods for zero-shot learning, such as 'Boosting Zero-shot Stereo Matching using Large-scale Mixed Images Sources in the Real World'. Overall, these emerging trends and techniques are expected to play a crucial role in the development of more accurate and reliable autonomous systems.
Emerging Trends in Autonomous Perception and Localization
Sources
Camera-Only Bird's Eye View Perception: A Neural Approach to LiDAR-Free Environmental Mapping for Autonomous Vehicles
RESAR-BEV: An Explainable Progressive Residual Autoregressive Approach for Camera-Radar Fusion in BEV Segmentation
Two-Stage Random Alternation Framework for Zero-Shot Pansharpening
Transformer-Based Dual-Optical Attention Fusion Crowd Head Point Counting and Localization Network
Technical Report for ICRA 2025 GOOSE 2D Semantic Segmentation Challenge: Leveraging Color Shift Correction, RoPE-Swin Backbone, and Quantile-based Label Denoising Strategy for Robust Outdoor Scene Understanding
DepthFusion: Depth-Aware Hybrid Feature Fusion for LiDAR-Camera 3D Object Detection
FD-RIO: Fast Dense Radar Inertial Odometry
MDF: Multi-Modal Data Fusion with CNN-Based Object Detection for Enhanced Indoor Localization Using LiDAR-SLAM
Boosting Zero-shot Stereo Matching using Large-scale Mixed Images Sources in the Real World
Crowd Scene Analysis using Deep Learning Techniques
VGC-RIO: A Tightly Integrated Radar-Inertial Odometry with Spatial Weighted Doppler Velocity and Local Geometric Constrained RCS Histograms
APR-Transformer: Initial Pose Estimation for Localization in Complex Environments through Absolute Pose Regression
MoRAL: Motion-aware Multi-Frame 4D Radar and LiDAR Fusion for Robust 3D Object Detection
Camera-Only 3D Panoptic Scene Completion for Autonomous Driving through Differentiable Object Shapes
Unsupervised Radar Point Cloud Enhancement via Arbitrary LiDAR Guided Diffusion Prior
Depth Anything with Any Prior