{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/active-neural-localization","title":"Active Neural Localization","arxiv_id":"1801.08214","date":"2018-01-24","proceeding":"ICLR 2018 1","authors":["Devendra Singh Chaplot","Emilio Parisotto","Ruslan Salakhutdinov"],"abstract":"Localization is the problem of estimating the location of an autonomous agent\nfrom an observation and a map of the environment. Traditional methods of\nlocalization, which filter the belief based on the observations, are\nsub-optimal in the number of steps required, as they do not decide the actions\ntaken by the agent. We propose \"Active Neural Localizer\", a fully\ndifferentiable neural network that learns to localize accurately and\nefficiently. The proposed model incorporates ideas of traditional\nfiltering-based localization methods, by using a structured belief of the state\nwith multiplicative interactions to propagate belief, and combines it with a\npolicy model to localize accurately while minimizing the number of steps\nrequired for localization. Active Neural Localizer is trained end-to-end with\nreinforcement learning. We use a variety of simulation environments for our\nexperiments which include random 2D mazes, random mazes in the Doom game engine\nand a photo-realistic environment in the Unreal game engine. The results on the\n2D environments show the effectiveness of the learned policy in an idealistic\nsetting while results on the 3D environments demonstrate the model's capability\nof learning the policy and perceptual model jointly from raw-pixel based RGB\nobservations. We also show that a model trained on random textures in the Doom\nenvironment generalizes well to a photo-realistic office space environment in\nthe Unreal engine.","url_abs":"http://arxiv.org/abs/1801.08214v1","url_pdf":"http://arxiv.org/pdf/1801.08214v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"active-neural-localization","repo_url":"https://github.com/devendrachaplot/Neural-Localization","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"unanswered"}}],"tasks":[{"task_slug":"fps-games","task_name":"FPS Games"},{"task_slug":"game-of-doom","task_name":"Game of Doom"},{"task_slug":"reinforcement-learning","task_name":"Reinforcement Learning"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1801.08214","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}