Modified interactive multiple model particle filter for terrain referenced navigation with classification error minimisation strategy

The authors propose an innovative solution to address challenges in terrain-referenced navigation (TRN). The suggested solution is the interactive multiple-model particle filter with a classification error minimisation strategy (IMM-CPF) based on decision theory. TRN is a technique that estimates position by comparing measured terrain altitude to the digital elevation model and critically depends on obtaining accurate altitude measurements. However, these measurements can be easily contaminated to not only from sensor errors but also from vegetation effects. The TRN measurement noise model is characterised as a multi-modal density, and it reveals an overlap between two density functions, with the mixture weight parameter varying based on surface environmental conditions. This variability can potentially degrade estimation accuracy. The proposed approach integrates truncated likelihoods into the mode estimation process to enhance mode estimation capability using a classification error minimisation strategy. The proposed strategy is based on decision theory and has been modified to be suited in the IMMPF form. The effectiveness of the proposed IMM-CPF method is verified through simulations conducted under diverse surface conditions, demonstrating significant improvements in estimation accuracy compared to conventional algorithms. Furthermore, the significance of this method is presented in terms of computational cost and robustness.