An overview of differentiable particle filters for data-adaptive sequential Bayesian inference

By approximating posterior distributions with weighted samples, particle filters (PFs) provide an efficient mechanism for solving non-linear sequential state estimation problems. While the effectiveness of particle filters has been recognised in various applications, the performance of particle filters relies on the knowledge of dynamic models and measurement models, and the construction of effective proposal distributions. An emerging trend in designing particle filters is the differentiable particle filters (DPFs). By constructing particle filters' components through neural networks and optimising them by gradient descent, differentiable particle filters are a promising computational tool to perform inference for sequence data in complex high-dimensional tasks such as vision-based robot localisation. In this paper, we provide a review of recent advances in differentiable particle filters and their applications. We place special emphasis on different design choices of key components of differentiable particle filters, including dynamic models, measurement models, proposal distributions, optimisation objectives, and differentiable resampling techniques.