Id: arxiv/2611.11521 Url: https://arxiv.org/abs/2611.11521 Tags: cs.RO, cs.AI, cs.LG, eess.SY, math.DS Score: 0 Title: K-PWM: Control-Oriented Structured Bilinear World Models under Partial Observation Domain: cs.RO Source: arxiv Authors: Santosh Mohan Rajkumar, Sriram Narayanan, Samuel E. Otto, Debdipta Goswami Reviews: 0 Subject: cs.RO Category: research Comments: 0 Disputed: false Division: cs Mentions: 0 Discussed: 0 Published: 2026-11-08 Page: https://delphi.science/r/arxiv/2611.11521 World models learn internal simulators of environment dynamics for prediction, planning, and control, yet learning them from limited, noisy, and partial observations remains challenging. While much progress has focused on pixel-based observations, many physical systems instead provide vector-valued measurements that reveal only part of the underlying state. We introduce K-PWM, a probabilistic world model for prediction and control under partial observation. K-PWM couples Koopman-bilinear latent dynamics with a nonlinear observation decoder and performs probabilistic latent-state inference…