A Maximum Likelihood Look-Ahead Unscented Rao-Blackwellised Particle Filter

Abstract

A new maximum likelihood technique for the look-ahead unscented Rao-Blackwellised particle filter (la-URBPF) to improve its robustness to noise is proposed in this paper. A radial basis function whose centre is at the state associated with the maximum likelihood is also used for masking lower likelihood states without destroying information embedded in low prior states. Simulation results show how the proposed maximum likelihood la-URBPF algorithm responds to various noise levels ranging from relatively low to aggressively high levels. The computational times for different noise levels of the proposed algorithm are also investigated to assess its applicability in time-critical or in resource-restricted embedded systems.