CR 17/01/2022
- two test cases: Zermelo + Goddard (+ state constraint in the two cases)
-
@jcaillau @x-ETrela direct code on Zermelo -
@jcaillau @x-ETrela direct code on Goddard -
@jcaillau @x-ETrela shooting code on Zermelo -
@jcaillau @x-ETrela shooting code on Goddard -
@x-HZidan HJB on Zermelo -
@x-HZidan HJB on Goddard - direct / shooting: Python or Julia code (easily reproducible)
- HJB: ROCHJ
- inward pointing constraint (implies normality - not applicable for Goddard or Zermelo)
- Goddard / Zermelo: regular / singular turnpike for large time
- deadline end of March 2022
- max. length 50 pages
- RL (reinforcement learning): link with HJB, section 5 / conclusion / remark
- emphasis on complementarity of methods, each having its strength
- typical scenario: direct code or (even better to have gain control on local minima) HJB to detect structure and initialise shooting (precise solving)
- next meeting: 10/02/2022, 16:00
References:
- Sanz-Serna review on direct method
- Hasnaa et al paper
- forward-backward sweeping (after Chernousko et al (1982)), recent ref.
- paper on Goddard
- paper on singular turnpike
Edited by CAILLAU Jean-Baptiste