mlsanityData to 7 Oct 2026

Deep Active Inference for Pixel-Based Discrete Control: Evaluation on the Car Racing Problem

arXiv 2109.04155 ↗First posted 9 Sep 2021cs.AI cs.NEcited by 2

Not on the map. Why?

Abstract (arXiv)

Despite the potential of active inference for visual-based control, learning the model and the preferences (priors) while interacting with the environment is challenging. Here, we study the performance of a deep active inference (dAIF) agent on OpenAI's car racing benchmark, where there is no access to the car's state. The agent learns to encode the world's state from high-dimensional input through unsupervised representation learning. State inference and control are learned end-to-end by optimizing the expected free energy. Results show that our model achieves comparable performance to deep Q-learning. However, vanilla dAIF does not reach state-of-the-art performance compared to other world model approaches. Hence, we discuss the current model implementation's limitations and potential architectures to overcome them.

Key facts. Cited by 2 papers in the map (data through 7 Oct 2026). Builds on 16 of its 27 references that are in the map. First posted on arXiv 9 Sep 2021.

Builds on

16 of 27 references.

References · 16

  1. 2018World Modelscited by 1,444
  2. 2020Active inference on discrete state-spaces: a synthesiscited by 91
  3. 2020Deep active inference agents using Monte-Carlo methodscited by 72
  4. 2019Weight Agnostic Neural Networkscited by 71
  5. 2019Deep Active Inference as Variational Policy Gradientscited by 52
  6. 2020Reinforcement Learning through Active Inferencecited by 47
  7. 2017Deep Active Inferencecited by 44
  8. 2019Scaling active inferencecited by 30
  9. 2019End-to-End Pixel-Based Deep Active Inference for Body Perception and Actioncited by 24
  10. 2019Deep Neuroevolution of Recurrent and Discrete World Modelscited by 22
  11. 2021Multimodal VAE Active Inference Controllercited by 21
  12. 2020Deep Active Inference for Partially Observable MDPscited by 20
4 more
  1. 2019Bayesian policy selection using active inferencecited by 12
  2. 2021Neuroscience-inspired perception-action in robotics: applying active inference for state estimation, control and self-perceptioncited by 5
  3. 2021Online reinforcement learning with sparse rewards through an active inference capsulecited by 3
  4. 2020Deep Reinforcement Learning with Mixed Convolutional Networkcited by 1

Built on by

Cited by year

Citing papers per year: 2021: 1; 2022: 0; 2023: 0; 2024: 1.

2 papers.

  1. 2021Active Inference in Robotics and Artificial Agents: Survey and Challengescited by 52
  2. 2024R-AIF: Solving Sparse-Reward Robotic Tasks from Pixels with Active Inference and World Modelscited by 10

Similar, not cited

Close in wording, no citation either way

  1. 2024Towards Human-Like Driving: Active Inference in Autonomous Vehicle Control
  2. 2025Real-World Robot Control by Deep Active Inference With a Temporally Hierarchical World Model
  3. 2021Latent Imagination Facilitates Zero-Shot Transfer in Autonomous Racing
  4. 2021Vision-Based Autonomous Car Racing Using Deep Imitative Reinforcement Learning
  5. 2024Self-Driving Car Racing: Application of Deep Reinforcement Learning
  6. 2025Perceptual Motor Learning with Active Inference Framework for Robust Lateral Control
  7. 2021ObserveNet Control: A Vision-Dynamics Learning Approach to Predictive Control in Autonomous Vehicles
  8. 2026Lagrange: An Open-Vocabulary, Energy-Based Sparse Framework for Generalized End-to-End Driving