arXiv:1606.05830v4 [cs.RO] 30 Jan 2017 g

1 This paper has been accepted for publication in IEEE Transactions on : Explore: cite the paper as:C. Cadena and L. Carlone and H. Carrillo and Y. Latif and D. Scaramuzza and J. Neira and I. Reid and Leonard, Past, Present, and Future of simultaneous localization And mapping : towards the Robust-Perception Age ,in IEEE Transactions on Robotics 32 (6) pp 1309-1332, 2016bibtex:@article{Cadena16tro-SLAM future,title={Past, Present, and Future of simultaneous localization And mapping : towards the Robust-Perception Age},author={C. Cadena and L. Carlone and H. Carrillo and Y. Latif and D. Scaramuzza and J. Neira and I. Reid and Leonard},journal={{IEEE Transactions on Robotics}},year={2016},number={6},pages= {1309 1332},volume={32}} [ ] 30 Jan 20171 Past, Present, and Future of SimultaneousLocalization And mapping : towards theRobust-Perception AgeCesar Cadena, Luca Carlone, Henry Carrillo, Yasir Latif,Davide Scaramuzza, Jos e Neira, Ian Reid, John J.

2 LeonardAbstract simultaneous localization And mapping (SLAM)consists in the concurrent construction of a model of theenvironment (themap), and the estimation of the state of the robotmoving within it. The SLAM community has made astonishingprogress over the last 30 years, enabling large-scale real-worldapplications, and witnessing a steady transition of this technologyto industry. We survey the current state of SLAM and considerfuture directions. We start by presenting what is now thede-factostandard formulation for SLAM. We then review related work,covering a broad set of topics including robustness and scalabilityin long-term mapping , metric and semantic representations formapping, theoretical performance guarantees, active SLAM andexploration, and other new frontiers.

3 This paper simultaneouslyserves as a position paper and tutorial to those who are users ofSLAM. By looking at the published research with a critical eye,we delineate open challenges and new research issues, that stilldeserve careful scientific investigation. The paper also containsthe authors take on two questions that often animate discussionsduring robotics conferences:Do robots need SLAM?andIs SLAM solved?Index Terms Robots, SLAM, localization , mapping , Factorgraphs, Maximum a posteriori estimation, sensing, material for this paper, including an ex-tended list of references (bibtex) and a table of point-ers to online datasets for SLAM, can be found Cadena is with the Autonomous Systems Lab, ETH Z urich, Carlone is with the Laboratory for Information and Decision Systems,Massachusetts Institute of Technology, USA.

4 Carrillo is with the Escuela de Ciencias Exactas e Ingenier a, UniversidadSergio Arboleda, Colombia, and Pontificia Universidad Javeriana, Latif and I. Reid are with the School of Computer Science, Universityof Adelaide, Australia, and the Australian Center for Robotic Vision. Neira is with the Departamento de Inform atica e Ingenier a de Sistemas,Universidad de Zaragoza, Spain. Scaramuzza is with the Robotics and Perception Group, University ofZ urich, Switzerland. Leonard is with Marine Robotics Group, Massachusetts Institute ofTechnology, USA. paper summarizes and extends the outcome of the workshop TheProblem of Mobile Sensors: Setting future goals and indicators of progress forSLAM [25], held during theRobotics: Science and System(RSS) conference(Rome, July 2015).

5 This work has been partially supported by the following grants: MINECO-FEDER DPI2015-68905-P, Grupo DGA T04-FSE; ARC grants DP130104413,CE140100016 and FL130100102; NCCR Robotics; PUJ 6601; EU-FP7-ICT-Project TRADR 609763, EU-H2020-688652 and INTRODUCTIONSLAM comprises the simultaneous estimation of the stateof a robot equipped with on-board sensors, and the con-struction of a model (themap) of the environment that thesensors are perceiving. In simple instances, the robot state isdescribed by its pose (position and orientation), although otherquantities may be included in the state, such as robot velocity,sensor biases, and calibration parameters. The map, on theother hand, is a representation of aspects of interest ( ,position of landmarks, obstacles) describing the environmentin which the robot need to use a map of the environment is , the map is often required to support other tasks; forinstance, a map can inform path planning or provide anintuitive visualization for a human operator.

6 Second, the mapallows limiting the error committed in estimating the state ofthe robot. In the absence of a map, dead-reckoning wouldquickly drift over time; on the other hand, using a map, ,a set of distinguishable landmarks, the robot can reset itslocalization error by re-visiting known areas (so-calledloopclosure). Therefore, SLAM finds applications in all scenariosin which a prior map is not available and needs to be some robotics applications the location of a set oflandmarks is knowna priori. For instance, a robot operating ona factory floor can be provided with a manually-built map ofartificial beacons in the environment. Another example is thecase in which the robot has access to GPS (the GPS satellitescan be considered as moving beacons at known locations).

7 Insuch scenarios, SLAM may not be required if localization canbe done reliably with respect to the known popularity of the SLAM problem is connected with theemergence of indoor applications of mobile robotics. Indooroperation rules out the use of GPS to bound the localizationerror; furthermore, SLAM provides an appealing alternativeto user-built maps, showing that robot operation is possible inthe absence of an ad hoc localization thorough historical review of the first 20 years of theSLAM problem is given by Durrant-Whyte and Bailey intwo surveys [7, 69]. These mainly cover what we call theclassical age(1986-2004); the classical age saw the intro-duction of the main probabilistic formulations for SLAM,including approaches based on Extended Kalman Filters, Rao-Blackwellised Particle Filters, and maximum likelihood esti-mation; moreover, it delineated the basic challenges connected2to efficiency and robust data association.

8 Two other excellentreferences describing the three main SLAM formulationsof the classical age are the book of Thrun, Burgard, andFox [240] and the chapter of Stachnisset al.[234, Ch. 46].The subsequent period is what we call thealgorithmic-analysisage(2004-2015), and is partially covered by [64]. The algorithmic analysis period saw the studyof fundamental properties of SLAM, including observability,convergence, and consistency. In this period, the key role ofsparsity towards efficient SLAM solvers was also understood,and the main open-source SLAM libraries were review the main SLAM surveys to date in Table I,observing that most recent surveys only cover specific aspectsor sub-fields of SLAM.

9 The popularity of SLAM in the last 30years is not surprising if one thinks about the manifold aspectsthat SLAM involves. At the lower level (called thefront-endin Section II) SLAM naturally intersects other research fieldssuch as computer vision and signal processing; at the higherlevel (that we later call theback-end), SLAM is an appealingmix of geometry, graph theory, optimization, and probabilisticestimation. Finally, a SLAM expert has to deal with practicalaspects ranging from sensor calibration to system present paper gives a broad overview of the current stateof SLAM, and offers the perspective of part of the communityon the open problems and future directions for the SLAM research.

10 Our main focus is on metric and semantic SLAM,and we refer the reader to the recent survey by Lowryetal.[160], which provides a comprehensive review of vision-based place recognition and topological delving into the paper, we first discuss two questionsthat often animate discussions during robotics conferences:(1) do autonomous robots need SLAM? and (2) is SLAM solved as an academic research endeavor? We will revisit thesequestions at the end of the the question Do autonomous robots really needSLAM? requires understanding what makes SLAM aims at building a globally consistent representationof the environment, leveraging both ego-motion measurementsand loop closures.