The field of autonomous systems and robotics is witnessing significant advancements in sensor fusion, localization, and mapping technologies, aimed at enhancing the perception and navigation capabilities of autonomous vehicles and drones. A notable trend is the integration of multiple sensor modalities, such as LiDAR, radar, and cameras, to overcome the limitations of individual sensors and improve the robustness and accuracy of environmental perception. This is particularly evident in the development of frameworks that leverage the complementary strengths of radar and LiDAR for 3D object detection, and the use of LiDAR data in conjunction with map priors for accurate localization in GNSS-denied environments.

Another key development is the focus on optimizing algorithms for real-time performance on resource-constrained platforms, such as FPGAs and edge computing devices. This includes the design of lightweight, fully quantized 3D object detection algorithms and efficient SLAM methods that reduce computational load and memory usage without significantly compromising accuracy.

In the realm of drone technology, there is a push towards enabling autonomous operations in challenging environments, such as dense forests, through the use of advanced robotics and photogrammetric methods. This includes the development of under-canopy drones capable of autonomous flight and accurate tree parameter estimation, which is crucial for applications like forest inventory.

Finally, the field is also exploring novel approaches to rigid body localization and map optimization that do not rely on traditional anchor-based methods or require extensive computational resources. These advancements are paving the way for more efficient and scalable solutions to the challenges of autonomous navigation and environmental mapping.

Noteworthy Papers

• MutualForce: Introduces a 4D radar-LiDAR framework that mutually enhances their representations, achieving superior object detection performance.
• Multi-LiCa: Presents a motion- and targetless approach for multi LiDAR-to-LiDAR calibration, offering a generalized solution for various sensor setups.
• OpenLiDARMap: Proposes a method for creating georeferenced maps without GNSS support, leveraging publicly available data to eliminate long-term drift.
• LiFT: Develops a lightweight, FPGA-tailored 3D object detection algorithm, optimizing for real-time inference with minimal computational complexity.
• Automatic Labelling & Semantic Segmentation with 4D Radar Tensors: Demonstrates an automatic labelling process and semantic segmentation network that significantly improves vehicle detection performance.
• Towards autonomous photogrammetric forest inventory: Advances the capability of under-canopy drones for autonomous forest data collection and accurate tree parameter estimation.
• Egoistic MDS-based Rigid Body Localization: Introduces a novel anchorless rigid body localization method suitable for autonomous driving applications.
• Grid-based Submap Joining: Offers an efficient algorithm for optimizing global occupancy maps and local submap frames simultaneously in large-scale environments.
• VIGS SLAM: Proposes a 3D Gaussian Splatting SLAM method that integrates IMU sensor measurements for large-scale indoor environments.
• GeomGS: Introduces a LiDAR-guided Geometry-Aware Gaussian Splatting method for improved robot localization and mapping.
• FAST-LIVO2 on Resource-Constrained Platforms: Presents a lightweight LiDAR-inertial-visual odometry system optimized for resource-constrained platforms, achieving significant reductions in runtime and memory usage.