The field of robotic perception and mapping is moving towards more robust and precise methods for scene understanding and object manipulation. Researchers are exploring new approaches to fuse different sensing modalities, such as visual and contact sensing, to improve the accuracy and reliability of robotic assembly and exploration tasks. One notable trend is the use of Gaussian processes and stochastic models to represent uncertainty and improve the efficiency of surface reconstruction and path planning algorithms. These advances have the potential to enable more autonomous and efficient robotic systems in various applications, including planetary exploration and object manipulation. Noteworthy papers in this area include:

• ContactFusion, which proposes a method for fusing visual and contact information to improve robotic assembly.
• Informative Path Planning, which evaluates an algorithm for mapping unknown planetary surfaces using Gaussian processes.
• Stochastic Poisson Surface Reconstruction, which presents a new method for reconstructing surfaces from oriented point clouds using geometric Gaussian processes.
• Visuo-Tactile Object Pose Estimation, which proposes a method for estimating object pose using visual and tactile information.