Learning to Navigate in Cities Without a Map

By Piotr Wojciech Mirowski, Matthew Koichi Grimes, Mateusz Malinowski, Karl Moritz Hermann, Keith Anderson, Denis Teplyashin, Karen Simonyan, Koray Kavukcuoglu, Andrew Zisserman, Raia Hadsell. Neural Information Processing Systems, 2018

347 citations recorded by the source corpus.

Navigating through unstructured environments is a basic capability of intelligent creatures, and thus is of fundamental interest in the study and development of artificial intelligence. Long-range navigation is a complex cognitive task that relies on developing an internal representation of space, grounded by recognisable landmarks and robust visual processing, that can simultaneously support continuous self-localisation ("I am here") and a representation of the goal ("I am going there"). Building upon recent research that applies deep reinforcement learning to maze navigation problems, we present an end-to-end deep reinforcement learning approach that can be applied on a city scale. Recognising that successful navigation relies on integration of general policies with locale-specific knowledge, we propose a dual pathway architecture that allows locale-specific features to be encapsulated, while still enabling transfer to multiple cities. We present an interactive navigation environment that uses Google StreetView for its photographic content and worldwide coverage, and demonstrate that our learning method allows agents to learn to navigate multiple cities and to traverse to target destinations that may be kilometres away. The project webpage this http URL contains a video summarising our research and showing the trained agent in diverse city environments and on the transfer task, the form to request the StreetLearn dataset and links to further resources. The StreetLearn environment code is available at this https URL

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Transportation

View Paper: https://www.semanticscholar.org/paper/86430b50921c7aecb212b4cd0a43a73b091cf95f

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

In Transportation

Citations in the corpus

Talk the Walk: Navigating New York City through Grounded Dialogue (2018, background): “Recently, there has been increased interest in using end-to-end trainable neural networks for learning to navigate indoor scenes(Gupta et al., 2017b;a) or large cities (Brahmbhatt & Hays, 2017; Mirowski et al., 2018), but, unlike our work, without multi-agent communication.”
Scene Representation Transformer: Geometry-Free Novel View Synthesis Through Set-Latent Scene Representations (2021, background): “Such a system can be useful for virtual exploration of urban spaces [26], as well as other mapping, visualization, and AR/VR applications [27].”
Learning to Visually Navigate in Photorealistic Environments Without any Supervision (2020, neutral): “[20] learn a navigation policy with a recurrent network in synthetic mazes, and later, in real-world data from Google Maps [19].”
Learning Deployable Navigation Policies at Kilometer Scale from a Single Traversal (2018, methodology): “We mitigate this problem by training the agent on a curriculum of gradually increasing difﬁculty [30, 31, 17].”
Self-Monitoring Navigation Agent via Auxiliary Progress Estimation (2019, result): “…(Das et al., 2018; Gordon et al., 2018), and navigation (Matuszek et al., 2013; Hemachandra et al., 2015; Duvallet et al., 2016; Zhu et al., 2017; de Vries et al., 2018; Yuke Zhu, 2017; Mousavian et al., 2018; Wayne et al., 2018; Wang et al., 2018a; Mirowski et al., 2017; 2018; Zamir et al., 2018).”
TOUCHDOWN: Natural Language Navigation and Spatial Reasoning in Visual Street Environments (2018, methodology): “In general, the performance illustrates the challenge of the task.”
On Evaluation of Embodied Navigation Agents (2018, background): “Discrete environments can be convenient and can support interesting experiments and results [1, 3, 19, 22], but continuous spaces better reﬂect the conditions in which agents will be deployed in the physical world.”
Neural Topological SLAM for Visual Navigation (2020, background): “This ranges from reinforcement learning with sparse and shaped rewards [24, 25, 31, 33, 47], imitation learning and DAgger [17, 32], self-supervised learning for individual components [15, 34].”
Transferable Representation Learning in Vision-and-Language Navigation (2019, background): “Embodied Agent Navigation Navigation in realistic 3D environments has also received increased interest recently [35, 18, 29, 46].”
Deep Reinforcement Learning (2018, background): “ven exploration by self-supervised prediction. Watch two talks,Efros(2017) andGupta(2017), about self-supervised learning. Goel et al.(2018) conduct unsupervised video object segmentation for deep RL.Mirowski et al. (2018) study learning to navigate in cities without a map.Hsu et al.(2018) study uns”
End-to-End Autonomous Driving: Challenges and Frontiers (2023, neutral): “…designing these differentiable perception modules, several questions arise regarding the choice of loss functions, such as the necessity of 3D bounding boxes for objectdetection,whetheroptingforBEVsegmentationoverlane topologyforstaticsceneperception,orthetrainingstrategies withlimitedmodules’data.”
REVERIE: Remote Embodied Visual Referring Expression in Real Indoor Environments (2019, methodology): “f existing simulators roughly fall within two categories: synthetic-photo simulator [2, 15, 27] or realistic-photo simulator [1, 23, 28]. These simulators have been employed in a range of VLN methods [7, 20, 25]. Wang et al. [25] for example propose a look-ahead module to take into account the futur”
Visual Representations for Semantic Target Driven Navigation (2018, background): “Reinforcement learning (RL) strategies and associated deep learning architectures are discussed by Mirowski et al. [4], [13].”
RoboTHOR: An Open Simulation-to-Real Embodied AI Platform (2020, background): “Navigation has also been explored in other contexts such as autonomous driving (e.g., [12, 73]) or city navigation (e.g., [38, 13]).”
Rapid Task-Solving in Novel Environments (2020, background): “In past work with StreetLearn, agents were faced with a single large map (e.g. a city), in which they could learn to navigate over billions of experiences (Mirowski et al., 2018).”
Bayesian Relational Memory for Semantic Visual Navigation (2019, methodology): “More recently, thanks to the advances of deep learning, end-to-end approaches have been applied to tackle navigation in various domains, such as mazes [36, 27], indoor scenes [63, 5, 48, 37, 30], autonomous driving [7, 59, 13], and Google street view [35].”
Learning to plan with uncertain topological maps (2020, neutral): “Other recent work uses street-view scenes to train an agent to navigate in city environments [35].”
A Cordial Sync: Going Beyond Marginal Policies for Multi-Agent Embodied Tasks (2020, background): “Embodied navigation also aids AI agents to develop behavior such as instruction following [38,4,82,95,3], city navigation [18,64,63,94], question answering [21,22,34,97,24], and active visual recognition [105,104].”
Retouchdown: Adding Touchdown to StreetLearn as a Shareable Resource for Language Grounding Tasks in Street View (2020, background): “In this context, StreetLearn (Mirowski et al., 2018, 2019) stands out as a publicly available resource of Street View data that has been approved for dissemination and use for academic research.”
Vision-Dialog Navigation by Exploring Cross-Modal Memory (2020, background): “[15, 22] However, the lack of photorealism and natural language instruction limits the application of these environments.”