Titolo della tesi: Continous Mapping: Towards Non-Rigid SLAM
As technology advances, automation continues to reshape industries, making autonomous
systems increasingly dependent on precise mapping for navigation and decision-making.
The effectiveness of these systems in complex and dynamic environments relies on ac-
curate spatial awareness and reliable localization. Mapping has progressed from basic
2D representations to high-resolution 3D models, offering a deeper understanding of the
surroundings. However, conventional Simultaneous Localization and Mapping (SLAM)
methods assume rigid transformations, leading to motion distortions that degrade mapping
accuracy in real-world applications. Addressing these challenges requires a transition toward
non-rigid SLAM and advanced sensor fusion methodologies.
This research introduces novel approaches to refine non-rigid SLAM by integrating
motion compensation algorithms, multi-sensor fusion, and dynamic trajectory optimiza-
tion. Unlike conventional methods, this framework continuously refines pose estimation by
correcting measurement distortions, ensuring accuracy and stability over time. By enhanc-
ing spatial coherence and reducing trajectory drift, non-rigid SLAM enables autonomous
systems to operate more reliably in dynamic, real-world conditions.
Recognizing the limitations of existing SLAM datasets, this research also presents a
comprehensive robotics perception dataset. It includes LiDAR point clouds, RGB imagery,
IMU, and GPS data collected across diverse environments, providing a robust benchmark
for assessing motion-aware SLAM strategies. The dataset emphasizes precise calibration
and synchronization, utilizing modern equipment to capture various environments. Data
collection is conducted both manually and through vehicle-mounted systems, ensuring its
applicability to various robotic applications.
In summary, non-rigid SLAM represents a significant step toward more adaptable and
resilient mapping systems. By addressing sensor motion distortions and refining pose
estimation, this research contributes to the evolution of autonomous navigation, where
continuous-motion SLAM improves spatial awareness and enhances operational reliability
across a diverse range of applications.