1: \begin{abstract}

2: Existing model-based reinforcement learning methods often study perception modeling and decision making separately.

3: We introduce joint Perception and Control as Inference (PCI),

4: a general framework to combine perception and control for partially observable environments through Bayesian inference.

5: Based on the fact that object-level inductive biases are critical in human perceptual learning and reasoning,

6: we propose Object-based Perception Control (OPC),

7: an instantiation of PCI which manages to facilitate control using automatic discovered object-based representations.

8: We develop an unsupervised end-to-end solution and analyze the convergence of the perception model update.

9: Experiments in a high-dimensional pixel environment demonstrate the learning effectiveness of our object-based perception control approach. 

10: Specifically, we show that OPC achieves good perceptual grouping quality and outperforms several strong baselines in accumulated rewards.

11: \end{abstract}

12: