Least-squares particle filter

A least-squares particle filter (LSPF) is proposed, where the latest measurement is effectively integrated into each sampled particle with the help of the least-squares estimate, thereby promoting the movement of particles from the prior areas to the high likelihood regions. More importantly, an approach of augmenting the measurement vector is presented to overcome the irreversible problem potentially existing in the proposed filter, thus expanding the application range of the LSPF. The experimental results of bearings-only tracking demonstrate the better estimation accuracy of the LSPF than the standard particle filter and the auxiliary particle filter.