Active SLAM Paper List

(This figure is referenced from PyPolo.)

The autonomous robotic system is a key objective pursued by many researchers in the field of robotics. In recent years, advancements in state estimators, motion planning algorithms, 3D representations, and even large language models (LLMs) have led to a surge of efforts aimed at achieving autonomy through Active SLAM. This approach has become a focal point for developing systems capable of navigating and interacting with complex environments independently.

This repository organizes relevant papers in the Active SLAM domain according to the author's perspective. It is important to note that the classification used in the repository reflects the author's viewpoint; in reality, many similar works are difficult to categorize satisfactorily.

Overview

• Survey Paper
• 1 Robotic Exploration
  • 1.1 Single Robot
  • 1.2 Multi-robot System
  • 1.3 Neural-based Exploration
  • 1.4 Reinforcement Learning-based Exploration
  • 1.5 Visual Language Navigation/Exploration
• 2 Active SLAM
• 3 Active Loop Closure
• 4 Active Reconstruction/Mapping
  • 4.1 Single Object
  • 4.2 Entire Scene
• 5 Active Localization
  • 5.1 Active Object/Source Localization
  • 5.2 Active Self-localization
• 6 Robotic/Active Information Gathering

Survey Paper

• Past, Present, and Future of Simultaneous Localization and Mapping: Toward the Robust-Perception Age, TRO 2016. [Paper]

• A survey on active simultaneous localization and mapping: State of the art and new frontiers, TRO 2023. [Paper]

• Active slam: A review on last decade, Sensors 2023. [Paper]

1 Robotic Exploration

1.1 Single Robot

• A frontier-based approach for autonomous exploration, CIRA 1997. [Paper]

• Active Monocular Localization: Towards Autonomous Monocular Exploration for Multirotor MAV, ICRA 2014. [Paper]

• Confidence-rich Localization and Mapping based on Particle Filter for Robotic Exploration, IROS 2022. [Paper]

• TARE: A Hierarchical Framework for Efficiently Exploring Complex 3D Environments, RSS 2021. [Paper]

• Estimating Map Completeness in Robot Exploration, arXiv 2024. [Paper]

• 3D Active Metric-Semantic SLAM, RAL 2024. [Paper] [Code]

1.2 Multi-robot System

• Frontier-Based Exploration for Multi-Robot Rendezvous in Communication-Restricted Unknown Environments, arXiv 2024. [Paper]

1.3 Neural-based Exploration

• Learning to plan in high dimensions via neural exploration-exploitation trees, ICLR 2020. [Paper]
• Neural Topological SLAM for Visual Navigation, CVPR 2020. [Paper] [Website]
• Object Goal Navigation using Goal-Oriented Semantic Exploration, NeurIPS 2020, [Paper] [Code]
• Learning to Explore using Active Neural SLAM, ICLR 2020. [Paper] [Code] [Website]
• DARE: Diffusion Policy for Autonomous Robot Exploration, arXiv 2024. [Paper]

1.4 Reinforcement Learning-based Exploration

• A deep reinforcement learning approach for active SLAM, Applied Sciences 2020. [Paper]

• IR2 : Implicit Rendezvous for Robotic Exploration Teams under Sparse Intermittent Connectivity, IROS 2024. [Paper] [Code]

• HDPlanner: Advancing Autonomous Deployments in Unknown Environments through Hierarchical Decision Networks, RAL 2024. [Paper] [Code]

• Privileged Reinforcement and Communication Learning for Distributed, Bandwidth-limited Multi-robot Exploration, arXiv 2024. [Paper]

• Deep Reinforcement Learning-based Large-scale Robot Exploration, RAL 2024. [Paper]

1.5 Visual Language Navigation/Exploration

• Vlfm: Vision-language frontier maps for zero-shot semantic navigation, ICRA 2024. [Paper] [Code] [Website]

2 Active SLAM

Note: It is important to note that many studies claim to be "Active SLAM," yet most of them are primarily focused on Robotics Exploration. We contend that a true Active SLAM approach should prioritize the accuracy of both localization and mapping, actively optimizing these aspects by incorporating mechanisms such as an active loop closure module.

• On the monotonicity of optimality criteria during exploration in active SLAM, ICRA 2015. [Paper]
• On the uncertainty in active slam: representation, propagation and monotonicity, Doctoral dissertation 2018. [Paper]
• On the importance of uncertainty representation in active SLAM, TRO 2018. [Paper]
• Active SLAM using Connectivity Graphs as Priors, IROS 2019. [Paper]
• Active SLAM using 3D Submap Saliency for Underwater Volumetric Exploration, ICRA 2020. [Paper]
• Active SLAM With Prior Topo-Metric Graph Starting At Uncertain Position, RAL 2021. [Paper]
• REAL: Rapid Exploration with Active Loop-Closing toward Large-Scale 3D Mapping using UAVs, IROS 2021. [Paper]
• ExplORB-SLAM: Active Visual SLAM Exploiting the Pose-graph Topology, Iberian Robotics conference, 2022. [Paper] [Code]
• Loop-Aware Exploration Graph: A concise representation of environments for exploration and active loop-closure, RAS 2022. [Paper]
• Graph-based SLAM-Aware Exploration with Prior Topo-Metric Information, RAL 2024. [Paper] [Code]

3 Active Loop Closure

• Active Loop Closure for OSM-guided Robotic Mapping in Large-Scale Urban Environments, IROS 2024. [Paper] [Code]

4 Active Reconstruction/Mapping

In practice, it is challenging to distinctly differentiate between Active Reconstruction/Mapping tasks and Robotic Exploration tasks, as illustrated in the diagram above. Both involve the intersection of planning and mapping; however, their focus has notable differences. Active Reconstruction/Mapping emphasizes the quality of reconstructing entire scenes or objects, typically utilizing 3D frontiers and viewpoints. In contrast, Robotic Exploration prioritizes rapid coverage and efficiency, often employing 2D frontiers to optimize speed.

4.1 Single Object

• Simultaneous localization and map-building using active vision, TPAMI, 2002. [Paper]

• Uncertainty guided policy for active robotic 3d reconstruction using neural radiance fields, RAL 2022. [Paper]

• ActiveRMAP: Radiance Field for Active Mapping And Planning, arXiv 2022. [Paper]

• Active Implicit Object Reconstruction using Uncertainty-guided Next-Best-View Optimization, RAL, 2023. [Paper] [Code]

• FisherRF: Active View Selection and Mapping with Radiance Fields Using Fisher Information, ECCV, 2024. [Paper] [Code] [Website]

4.2 Entire Scene

• HGS-Planner: Hierarchical Planning Framework for Active Scene Reconstruction Using 3D Gaussian Splatting, arXiv, 2024. [Paper]

• Active Neural Mapping at Scale, arXiv, 2024. [Paper]

• ActiveSplat: High-Fidelity Scene Reconstruction through Active Gaussian Splatting, arXiv, 2024. [Paper] [Website]

5 Active Localization

5.1 Active Object/Source Localization

• Dual-stage planner for autonomous radioactive source localization in unknown environments, RAS 2024. [Paper]
• An Exploration-Enhanced Search Algorithm for Robot Indoor Source Searching, TRO 2024. [Paper]

5.2 Active Self-localization

6 Robotic/Active Information Gathering

• AK: Attentive Kernel for Information Gathering, RSS 2022. [Paper] [Code] [Website]

• Adaptive Robotic Information Gathering via Non-Stationary Gaussian Processes, IJRR 2024. [Paper] [PyPolo]