An adaptive data association for robotic SLAM in search and rescue operation

Data Association has been considered a difficult task in robotic SLAM (simultaneous localization and mapping), especially operating in deformed and unstructured environment with time variant objects such as earthquake shattered ruins. This paper proposes a SLAM data association and feature classification algorithm to identify the useful geometric features as landmarks even in case of deformation, by adapting the most likely discriminant function based on Bayesian analysis learning and a linear discriminant classification which works with any feature-based SLAM such as various Kalman filters, and particle filters. Due to the cost effectiveness in computation and complexity, this method can be applied for real-time SLAM applications. Simulation is performed to verify the effectiveness of method.