SACBP: Belief Space Planning for Continuous-Time Dynamical Systems via Stochastic Sequential Action Control

@inproceedings{morales_sacbp_2020,
  address = {Cham},
  title = {{SACBP}: {Belief} {Space} {Planning} for {Continuous}-{Time} {Dynamical} {Systems} via {Stochastic} {Sequential} {Action} {Control}},
  volume = {14},
  isbn = {978-3-030-44050-3 978-3-030-44051-0},
  shorttitle = {{SACBP}},
  url = {http://link.springer.com/10.1007/978-3-030-44051-0_16},
  abstract = {We propose a novel belief space planning technique for continuous dynamics by viewing the belief system as a hybrid dynamical system with time-driven switching. Our approach is based on the perturbation theory of diﬀerential equations and extends Sequential Action Control [1] to stochastic belief dynamics. The resulting algorithm, which we name SACBP, does not require discretization of spaces or time and synthesizes control signals in near real-time. SACBP is an anytime algorithm that can handle general parametric Bayesian ﬁlters under certain assumptions. We demonstrate the eﬀectiveness of our approach in an active sensing scenario and a model-based Bayesian reinforcement learning problem. In these challenging problems, we show that the algorithm signiﬁcantly outperforms other existing solution techniques including approximate dynamic programming and local trajectory optimization.},
  language = {en},
  urldate = {2020-09-15},
  booktitle = {Algorithmic {Foundations} of {Robotics} {XIII}},
  publisher = {Springer International Publishing},
  author = {Nishimura, Haruki and Schwager, Mac},
  editor = {Morales, Marco and Tapia, Lydia and Sánchez-Ante, Gildardo and Hutchinson, Seth},
  year = {2020},
  keywords = {belief\_space\_control},
  pages = {267--283}
}