A Structured Approach for Uncertain Transformations Trees

Kurzfassung

In the field of robotics, ensuring precise representation of spatial transformations is imperative for maintaining reliable system performance. However, conventional approaches often prove inadequate due to their failure to consider internal inaccuracies in the robot and environmental factors. In the context of robotic systems, deviations from nominal transformations arise from various sources such as sensor decalibration, inaccuracies in joint positions, deformations induced by mechanical stress, and gravitational influences, among other contributing factors. The same applies to environmental uncertainties, where the registered poses of objects and landmarks suffer from limitations in the perception methods. This paper advocates for a paradigm shift by introducing a framework that incorporates uncertainty into transformation trees, utilizing Lie Algebra for a consistent computation. Our approach addresses the aforementioned challenges, providing a realistic and robust representation of transformations. We demonstrate the applicability and efficacy of our framework through real-world examples.