slr-ltlm-mr

A Systematic Literature Review on Long-Term Localization and Mapping for Mobile Robots

• Methodology
  • Identification
  • Screening
  • Quality Evaluation
  • Data Extraction
• Contacts
• Acknowledgements
• Funding

Methodology

Identification

Base string:

(robot* OR vehicle*) AND ((locali* AND map*) OR "slam") AND ("long term" OR "life long" OR lifelong)

Inquiries: 2160 records

• ACM Digital Library
  • 125 results
  • Title + abstract + author keywords
  • Title:((robot* OR vehicle*) AND ((locali* AND map*) OR "slam") AND ("long term" OR "life long" OR lifelong)) OR Abstract:((robot* OR vehicle*) AND ((locali* AND map*) OR "slam") AND ("long term" OR "life long" OR lifelong)) OR Keyword:((robot* OR vehicle*) AND ((locali* AND map*) OR "slam") AND ("long term" OR "life long" OR lifelong))
• Dimensions
  • 430 results
  • Title + abstract
  • (robot OR robots OR robotics OR vehicle OR vehicles) AND (((localize OR localization OR localizing OR localise OR localisation OR localising) AND (map OR maps OR mapping)) OR "slam") AND ("long term" OR "life long" OR lifelong)
• IEEE
  • 260 results (after removing 173 duplicate records of the inquiries using a single field)
  • IEEE limits the maximum number of wildcards (e.g., *) in a query to 7
  • Search results obtained using single queries for each search field wanted
  • Title
    • 6 results
    • ("Document Title":robot* OR "Document Title":vehicle*) AND (("Document Title":locali* AND "Document Title":map*) OR "Document Title":"slam") AND ("Document Title":"long term" OR "Document Title":"life long" OR "Document Title":lifelong)
  • Abstract
    • 193 results
    • ("Abstract":robot* OR "Abstract":vehicle*) AND (("Abstract":locali* AND "Abstract":map*) OR "Abstract":"slam") AND ("Abstract":"long term" OR "Abstract":"life long" OR "Abstract":lifelong)
  • Index terms
    • 158 results
    • ("Index Terms":robot* OR "Index Terms":vehicle*) AND (("Index Terms":locali* AND "Index Terms":map*) OR "Index Terms":"slam") AND ("Index Terms":"long term" OR "Index Terms":"life long" OR "Index Terms":lifelong)
  • Author keywords
    • 2 results
    • ("Author Keywords":robot* OR "Author Keywords":vehicle*) AND (("Author Keywords":locali* AND "Author Keywords":map*) OR "Author Keywords":"slam") AND ("Author Keywords":"long term" OR "Author Keywords":"life long" OR "Author Keywords":lifelong)
  • IEEE terms
    • 0 results
    • Query: ("IEEE Terms":robot* OR "IEEE Terms":vehicle*) AND (("IEEE Terms":locali* AND "IEEE Terms":map*) OR "IEEE Terms":"slam") AND ("IEEE Terms":"long term" OR "IEEE Terms":"life long" OR "IEEE Terms":lifelong)
  • INSPEC controlled terms
    • 1 result
    • ("INSPEC Controlled Terms":robot* OR "INSPEC Controlled Terms":vehicle*) AND (("INSPEC Controlled Terms":locali* AND "INSPEC Controlled Terms":map*) OR "INSPEC Controlled Terms":"slam") AND ("INSPEC Controlled Terms":"long term" OR "INSPEC Controlled Terms":"life long" OR "INSPEC Controlled Terms":lifelong)
  • INSPEC non-controlled terms
    • 73 results
    • ("INSPEC Non-Controlled Terms":robot* OR "INSPEC Non-Controlled Terms":vehicle*) AND (("INSPEC Non-Controlled Terms":locali* AND "INSPEC Non-Controlled Terms":map*) OR "INSPEC Non-Controlled Terms":"slam") AND ("INSPEC Non-Controlled Terms":"long term" OR "INSPEC Non-Controlled Terms":"life long" OR "INSPEC Non-Controlled Terms":lifelong)
• INSPEC (link maybe be broken due to its length, restart your INSPEC session if needed)
  • 428 results
  • Subject/Title/Abstract: equivalent to retrieving results from any of the fields abstract, title, controlled term, or uncontrolled term (link)
  • (((robot* OR vehicle*) AND ((locali* AND map*) OR "slam") AND ("long term" OR "life long" OR lifelong)) WN KY)
• Scopus
  • 530 results
  • TITLE-ABS-KEY: equivalent to retrieving results from any of the fields abstract, title, and keywords (link
  • TITLE-ABS-KEY((robot* OR vehicle*) AND ((locali* AND map*) OR "slam") AND ("long term" OR "life long" OR lifelong))
• Web of Science
  • 387 results
  • Topic: equivalent to retrieving results from any of the fields title, abstract, author keywords, and Keywords Plus
  • Query: TS=((robot* OR vehicle*) AND ((locali* AND map*) OR "slam") AND ("long term" OR "life long" OR lifelong))

Last inquiry: May 17, 2022

Eligibility criteria (exclusion):

• [E1] Index: Papers not indexed in a scientific publication venue
• [E2] Language: Full-text of the papers not published in English
• [E3] Subject Area: Papers not classified in the databases as Computer Science, Engineering, Mathematics, or Multidisciplinary
• [E4] Short Papers: Papers classified as short papers accordingly to the publication venue
• [E5] Gray, Secondary, and Tertiary Literature: Books, preprints, reports, reviews, thesis, ...
• [E6] Availability: Full-text of the papers not available in digital libraries
• [E7] Dataset: Papers that focus only on data collection
• [E8] Coverage: Papers using only odometry for localization
• [E9] Scope Papers: that focus on different and not related subjects

Records removed before screening:

• Duplicates: 1339 records
• Exclusion criteria: 232 records

Records identified from the search results: 589 records

Further information relative to this phase of the review can be found in identification/. The source code for checking if the references have all required information (while normalizing the BibTex files generated from the different data sources considered in the review) and to identify short papers is available in check_references/ and check_short_papers/, respectively.

Screening

Records screened: 589 records

Records removed after screening:

• Duplicates: 13 records
• Exclusion criteria: 165 records

Records accepted for quality assessment: 411 records

Further information relative to this phase of the review can be found in screening/.

Quality Evaluation

Records assessed for eligibility: 411 records

Quality evaluation criteria:

• [QE1] Does the paper have an updated state of the art on long-term localization and mapping? 0.0, 0.5, 1.0
• [QE2] Is the methodology appropriate and detailed? 0.0, 0.5, 1.0
• [QE3] Does the methodology consider both localization and mapping problems? 0.0, 1.0, 2.0
• [QE4 Is the hardware and/or software used in the experiments detailed? 0.0, 0.5, 1.0
• [QE5] Does the paper presents any kind of long-term experimental results? 0.0, 2.0
• [QE6] Does the paper presents comparative results with other methods and/or ground-truth data? 0.0, 1.0
• [QE7] Does the work's implementation and/or the data used in the experiments are publicly available? 0.0, 1.0
• [QE8] Is the discussion of the results and conclusions appropriate and detailed? 0.0, 0.5, 1.0

Records removed after quality assessment:

1. QE2, QE3, QE5, QE8 (= 0): 116 records
2. Cut-off score (<= 7.5/10.0): 120 records

Records accepted for analysis: 175 records

Further information relative to this phase of the review can be found in quality-assessment/.

Data Extraction

Records analyzed for data extraction: 175 records

Records removed for having extended versions: 33 records

Records to include in the review: 142 records

Data extraction items:

1. [DE1] Long-term considerations: long-term factors the works consider in their proposed approach and experiments
   • appearance - varying conditions, appearance changes, ...
   • dynamics - environment dynamics, dynamic elements, ...
   • sparsity - map pruning, redundant data removal, ...
   • multi-session - map management, ...
   • computational - memory management, efficiency, ...
2. [DE2] Localization: how the robot localizes itself and the type of localizer: Bayesian, particle filter, feature matching, point cloud matching, ...
3. [DE3] Mapping: type of the map. focused on how new information is added to the map the type of information: grid, features, topological (graph SLAM, experience maps), database (images, global features, ...), ...
4. [DE4] Multi-robot: if the proposed methodologies consider multi-robot systems: yes / no
5. [DE5] Execution mode: offline, online, if requires both, or if no information on this item: if "online", assumed that it is able to execute the algorithm both online and offline independently; "both" means that the method requires an offline computation and also an online one (dependent on each other)
6. [DE6] Environment and domain: type of environment (indoor, outdoor) and domains (air, ground, water) tested with the proposed methodologies
7. [DE7] Sensory setup: which sensors considered in the methodologies: wheel odometry, IMU, laser (2D/3D), camera (monocular / stereo / omnidirectional), sonar, radar, GPS, or a combination of the previous sensors
8. [DE8] Non-public experiments: if the authors performed experiments or tests with non-public data
9. [DE9] Ground-truth: how ground-truth for non-public data is obtained or its type, if available: external tracking system, GPS, SLAM-based, manual (annotation, labeling, manual correction, measures, ...)
10. [DE10] Distance and time characteristics: relative to the non-public experiments if available, as follows
    • total distance (km) of the non-public experiments
    • path (km), in the case of repetitive paths
    • total time (h) in terms of continuous operation
    • time interval (day/week/month/year, or d/w/m/y) between the first and the last run
11. [DE11] Datasets: if and which public datasets are used in the experiments: KITTI, Oxford RobotCar, Nordlandsbeen, MIT Stata Center, Intel
12. [DE12] Evaluation metrics: which metrics are used for evaluation: average traveled error (ATE), pose error (along time, mean, std), execution time, memory, precision-recall, f-score, confusion matrix...

Further information relative to this phase of the review can be found in data-extraction/.

Results Overview

All results retrieved from the bibliographic data of the 142 records included in the review can be found in results. The tool VOSviewer used for obtaining the co-occurrence analysis of keywords and co-authorships is available in its website or in src. Lastly, the thesaurus used for analyzing keywords co-occurrence can be found in VOSviewer/data.

Keywords co-occurrence

Original:

With thesaurus:

Co-authorship analysis

Publication year

Contacts

If you have any questions or you want to know more about the work developed by us, please contact one of the contributors of this project:

• Ricardo B. Sousa (gitlab, github, personal, feup:professor, feup:student, inesctec) (corresponding author)
• Héber Miguel Sobreira (gitlab, inesctec)
• António Paulo Moreira (feup)

Project Link: https://github.com/sousarbarb/slr-ltlm-mr

Acknowledgements

• CRIIS - Centre for Robotics in Industry and Intelligent Systems from INESC TEC - Institute for Systems and Computer Engineering, Technology and Science
• Faculty of Engineering, University of Porto (FEUP)

Funding

This work is financed by National Funds through the Portuguese funding agency, FCT - Fundação para a Ciência e a Tecnologia, within project LA/P/0063/2020. This work is also financed by National Funds through the Portuguese funding agency, FCT - Fundação para a Ciência e a Tecnologia, within scholarship 2021.04591.BD.