An ε \varepsilon -Optimal Sequential Approach for Solving zs-POSGs
Jilles S. Dibangoye, Matthia Sabatelli, Erwan C. Escudie · 2026
Read An ε \varepsilon -Optimal Sequential Approach for Solving zs-POSGs by Jilles S. Dibangoye, Matthia Sabatelli, Erwan C. Escudie (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-24092.txt.
Description
Erwan C. Escudie and two co-authors present an epsilon-optimal sequential approach for solving zero-sum partially observable stochastic games. The 2026 paper tackles a known barrier: reductions to transition-independent games admit dynamic programming in theory, but the simultaneous minimax backup is nonlinear and exponentially complex.
The authors recast the simultaneous interaction as a sequential decision process through a separation principle. Distinct sufficient statistics for valuation and execution, the sequential occupancy state and the private occupancy family, reveal latent geometry in the optimal value function. That structure linearizes the backup operator, cutting update complexity from exponential to polynomial. In short, a hard update turns into a simple one. The simple update still finds safe policies. That swap is what makes the hard games solvable. The team proves it on tests. That swap is the whole trick. Tests show that algorithms built on the sequential framework beat leading methods.
Catalog record
| CREATOR | Jilles S. Dibangoye, Matthia Sabatelli, Erwan C. Escudie |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 129 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-24092