Simultaneous Localization And Mapping (SLAM): A Comprehensive Guide

What is Simultaneous Localization and Mapping (SLAM)?

Question:

Discuss the basic apprehension regarding the Simultaneous Localization and mapping or SLAM.
 

The purpose of this article is to discuss the basic apprehension regarding the Simultaneous Localization and mapping or SLAM. This article also delves into the extensive researches that have been performed on Simultaneous Localization and mapping over the decades since, the time it was first devised (Ahn, Doh and Chung, 2015). Simultaneous Localization and Mapping is the system by the implementation of which a mobile automatic vehicle or robot is made to create a map of the environment where that mobile automatic vehicle or robot is currently located, and computes its own location on the map that it is creating at the same time. A rapid and intense development in this particular field has been observed in the past decade along with numerous compelling implementations in the use of SLAM methods.

SLAM an acronym for Simultaneous Localization and Mapping refers to a robotic mapping system. The SLAM is actually a computational algorithm used for updating or constructing a map of a location which is unknown to the user who is constructing the algorithm, aiming to keep track of the location of an agent within the map. This might seem as a problem that circulates on the same topic simultaneously, the fundamental purpose of SLAM technology is used widely mapping systems, where it is exclusively used to construct a map of an unknown location (Buck, 2014). For instance, if a group of researchers are exploring an unexplored area under the ocean, which might be too risky for them to approach directly; they deploy a probe i.e. a machine functioning on the principles of robotics, implement slam techniques in the robot/probe to explore the region which cannot be accessed by humans. In this example the robot/probe deployed by the group of researchers is the agent and the technology used for mapping the unknown location chosen for the exploration is SLAM or Simultaneous Localization and Mapping technique.

The technique is to solve a problem by means of making use of several algorithms. These algorithms are exclusively designed for the purpose of solving a SLAM problem so that the technique can be used accordingly (Carmack, 2014). Although there is no certainty or guarantee that a certain use of the SLAM technique will result in meeting the expectations however, in is assured that a certain level of success can be achieved while implementing slam techniques in the exploration and mapping purposes. This is due to the fact that the algorithms that are used to work out the SLAM problems have been designed to utilize the available resources and when mapping unknown regions it is capable of making use of the existing resources and create a map according to the resources that were previously incorporated in it. Furthermore, in order to obtain the approximate results proper filters are required to be implemented in solving the SLAM problem. In such cases filters like the Kalman Filter and Particle filter are found to be very effective.

The Development of SLAM Technology

Slam techniques and tested approaches are commonly used in domestic robots, unmanned airborne vehicles, auto functioning land vehicles, planetary rovers, self guiding underwater vehicles, and even exploration of the human body where most definitely no other means of exploration can be implemented (Corbett, n.d.).

The fundamental question that arises from the SLAM problem is that whether it is possible for the automatic means that is being used for the purpose, to be actually placed at an anonymous location and in an anonymous atmosphere. Furthermore, it also delves into the possibilities that whether it would be actually possible for that automatic means to create map of the location and further, compute back its own location in the map which it was meant to create in the first place. The problem seems critical and the possibilities of success are limited (Crew, Phanavanh and Garcia-Borges, 2014). Therefore, in a situation like this the solution for this problem has been compared with the Holy Grail in the mobile robotics vista. Since, implementation of a technology like this would result in giving autonomous capabilities to a robot or any other automatic means or vehicles. Therefore, it would open the window for new possibilities in the field of science and technology if achieved.

The solutions that were developed over the years to solve the problems relating to SLAM, has been one of the major breakthroughs in the field of technology in the past decade. The solution to the problem came from formulating the SLAM problem into a theoretical problem and then solving it in multiple forms so that; once the problem is deduced it would be unwise not to open access for multiple possibilities (Ferreira et al., 2013). After solving the SLAM problem, the approach was implemented in multiple domains to understand and analyze the end results that would come out from its implementation on robots, airborne and underwater systems before the concept could be used in the planetary rovers.

Conceptually and theoretically, SLAM at present can be considered to be a problem whose solution has been discovered. Yet, some of the significant issues still persist. These issues persist in comprehending or finding more generic solutions for SLAM, which would especially be effective in perceiving the surrounding environment and creating better maps as a component of the algorithm it is responsible to. This would not only enhance the functionality of the object on which the SLAM has been implemented upon but also aid in the creation of better and detailed maps of the region aimed for the purpose (Gentile, 2013). The progress that the modern technology is making towards this concept has been seen in the Mars Rover Project. The Mars Rover Project concept was based on the concept of SLAM techniques and uses the theories and formulas that resulted in giving reality to the SLAM concept. However, the method of SLAM incorporated in the Mars Rover is different than the method used in domestic and outdoor robots. The method of SLAM used in the Mars Rover is far superior and relies on the concept and theories mentioned at the start of this passage. 

SLAM Applications

The birth of the probability based SLAM problem happened in the year of 1986 at the IEEE Automation and Robotics symposium held in San Francisco. This was the period when the probability based methods were being used introduced in giving a conscience to the AI and robots. However, this period was just the beginning of the phase and the researchers associated with this vision were looking forward to apply theoretic estimation methods to solve localization and mapping predicaments. These group of researchers included Jim Crowley, Hugh Durrant Whyte and Peter Cheeseman (Havangi, 2015). The result of the symposium came out with the recognition given to the method or to better say to the concept of probabilistic localization and mapping. This was based on Durrant-Whyte, Cheeseman and Smith’s statistical work on the manipulation of geometric uncertainty between landmarks and describing the relationship between. The key feature of this statistical work was to demonstrate the requirement for a high degree correlation to be present between that which was to be the estimated location of the various landmarks in the map. In addition, it required to be focused on in a regard that the correlations would develop with each successive and subsequent observations (He et al., 2013).

During this period two other researchers dealing with the incorporation of probability based method in localization and mapping, Faugeras and Ayache had taken the responsibility in developing means that would provide visual navigation. Crowley with two other researchers Laumond and Chatila were concentrating on developing navigation systems for mobile robots that would incorporate sonar based approach by means of using Kalman’s Filter and that sort of algorithms (Kim, 2012). Both the research groups and the research works they were engaged in were more or less engaged in similar sort of work. In this period Cheeseman, Durrant-Whyte and Smith had obtained a breakthrough with their new theory.

This theory described that with every movement that the mobile robot makes in an unknown atmosphere, it will keep a track of its surrounding atmosphere. It will track and compute its observations of the surrounding landmarks along with its own location. The estimates of the landmarks were going to be precisely correlated since; the estimated vehicle location system had its errors. This suggestion that their theory offered was insightful (Köster, 2014). This would require a dependable and complete solution for the combined mapping and localization problem. Furthermore, this would require a joint status of the vehicle composing of its position and estimation of each of the landmark position, and this would be updated with the observation of each landmark the mobile robot will be passing through. This would apparently mean the robot will be computing its own position at a stipulated interval of time or on the basis of landmarks it came across. Moreover, it would require the one observing and noting down the estimates in the robots programming to make use of a massive state vector, which would include an order based on the numeric data representation on the landmarks which were being maintained in the particular map, complete with appropriate computing and scaling of the squares of the figure of landmarks (Milstein, 2012).

Algorithms Used in SLAM

Importantly enough this work did overlook the aspects of converging properties of the steady behavior of the map. At that point of time it was widely presumed that the errors consisting in the estimated map would less likely converge and instead showcase an error growth that was unbounded and random walk behavior. Therefore, provided with the complexity in the computing process regarding the predicament in the mapping process and the lack of knowledge regarding the converging behavior with the map, researchers now concentrated on finding approximations that would help them solve the constant predicament relating to mapping (Oh and Sim, 2014). This resulted or to better say forced the estimated correlations between the estimated landmarks to be eliminated or reduced depending the complexity as a result, reducing the entire filter to a series of landmarks that were decoupled to vehicle filters. In addition, as a consequence the theoretical exercise on the combined mapping and localization problem came to a momentary halt, where mapping and localization were dealt in separately with the objective to find separate solutions for each of them. This however, did not render the research to a permanent halt only hindered its progress for the time being only (Rethi, 2016).

It is apparent from the analysis on the history that the method dealt with an approach towards the problem which was of complex nature. Furthermore, the entire concept for the project was dealing with ascertaining the uncertain itself thus, it was bound to be hindered in some way or the other, and this was the momentary halt that the researchers of this project found themselves in. however, the conceptual advance came up when the researchers realized that the combination of localization and mapping problem that they were facing at the moment was the consequence had formulated from a single error in their estimation process, and it related to the earlier predicament of convergence (Rich, 2015). Importantly enough the researchers realized that the predicament was existing among the correlations of the landmarks. Most of the researchers had tried to solve this predicament but neglected the fact that this was the most crucial part. It was not late when the researchers realized the fact that the more the correlations sprung the better the solutions were to be discovered.  This was the answer that the researchers were looking forth to obtain from the questions that arose from past errors, complexities and failures. In the year of 1995 the combined efforts of these researchers and the breakthrough that they had achieved in the field mapping and localization estimation was presented in the Mobile Robotic Survey Paper in that very years International Symposium on Robotics Research (Schüz, 2012). It was this very moment that the term SLAM i.e. Simultaneous Localization and Mapping was coined along with its structure and a whole new theory relating to convergence was published.

At present a number of researches are being carried on based on the essential principles of this theory and dynamics on the field of robotics. In these researches the exponential possibilities of the SLAM technique is being implemented upon. Testing the concept on domestic and surveillance bots are being carried out to learn the various probabilities and possibilities the concept is capable of unlocking (Sowter, Hale and Starr, 2013). Some of the instances of the successes that these researches have achieved over the past decades can be seen in the robots, unmanned airborne vehicles, auto functioning land vehicles, planetary rovers, self guiding underwater vehicles, and even exploration of the human body where most definitely no other means of exploration can be implemented. 

Issues and Solutions in SLAM

The above sections dealt with the nature of the SLAM technique and the potential it has in the field of robotics, and using the robots for mapping and localization. The purpose of this section is to provide a critical analysis on the SLAM technique and appreciate the fact that it is indeed a major breakthrough that the 21^st century observed.

The orientation of the researchers when they had merely conceived the concept was to evaluate a means of estimation that will redefine exploration. Exploration back then was the only means through which the mapping and localization could be accomplished. With this conception in their minds they used mathematical probability and converted the probability into theoretical stratagem (Szewczyk, Zieliński and Kaliczyńska, n.d.). This conversion of mathematical formulae into a theoretical stratagem instilled in their minds a better understanding of actual matters that they had to deal with. With this comprehension on the goals they had to accomplish with the understanding they had perceived they set on to try implementing this understanding on different projects. The end results of the projects they had work on with this particular theory they were able to get a clear picture of what they would be required to do in order to give reality to their original concept.

As already discussed in its brief history the more complex the nature of the analysis was the better the solutions the researchers could obtain. This indeed was a distinguishing aspect of the concept. Since, most the theories in the field of technology focused more on minimizing the growing complex in the implementation of various concepts. This particular concept need not require control in this aspect (Taibi, 2014). Furthermore, the basis of the concept was the inclusion of more and more estimations in the surroundings of its agents. The more variables in the end results demonstrated the higher success rate.

In addition, the concept had just opened new boundaries for the extra-terrestrial exploration purposes. Researchers lost no time in capitalizing on this exposure, which in turn resulted in the development of the Mars rover project. This project focused on exploring the surface of the Mars which could not be accessed by normal means. This further required a method through which the probe or robot would be able to explore the surface of Mars and also compute its observation and location at the same time. However, the concept was implemented in the mars rover project that opened an all new possibility in this project (Williams et al., 2013).

Apart from that the SLAM technique was being implemented on drones that would operate automatically or via remote guidance system. This research also gained success as the drone would operate automatically in the environment which has been selected for its flight course, and collect observation of its nearest landmarks and constantly compute its own location.

The possibilities that the SLAM technique had just opened for the purpose of exploration, mapping and pinpointing localization were boundless. As, the operation of the concept required estimating the probabilities and variables that would greatly affect the possibilities of the end result, a few resources were required to carry out an operation based on this concept. Furthermore, the operations based on this concept did not necessarily require additional resources from the region chosen for the exploration, mapping and localization, researchers could just use the resources that they possessed in order to commence and carry out the operation (Williams and Canedy, 2012). However, there was and is still a drawback of the SLAM technique. It is that the technique is based on approximates and the results that are produced from the technique is based on approximates only. There is no assurance that the analysis could actually produce precise results. The concept of this technique is to provide an apprehension of the environment it is exploring and not a precise figure. Therefore, the results may vary upon further using mapping and localization technologies like using surveillance cameras.

SLAM in the Mars Rover Project

One of the prime need in which SLAM is used is that it has great efficiency when it comes to measurement. Measuring a certain distance between two points or the overall size of a certain region can be quite easily accomplished with the implementation of SLAM techniques. This procedure includes that the unmanned device or robot is set to cover a certain distance between two points, regardless of the way it has been directed to, the unmanned device or the robot will make the necessary observations around it and estimate the distance that it traversed during the operations. The most usual form that is used in measuring is the laser scanner like LIDAR. These type scanners incorporate laser in them and are far easy to use than most other means and offer accurate results (YUAN and ZHAO, 2011). Nevertheless, these are not known to be cheap therefore, widespread implementation of these scanners in measuring is not quite possible. Still, other options exist, e.g. the prototype that Crowley, and two of his fellow researchers worked upon used SONAR technology. This implementation of SONAR to the SLAM was found to be very effective and the end results were very promising. Furthermore, these technologies are far cheaper compared to laser scanners and can be afforded by researchers having small to moderate grants. In addition, using SONAR technology would help in mapping environments that are located underwater. Imaging devices can also be for mapping and localization purposes. These devices or optical readers as per their technical names are available in both 3d and 2d formats. The dependency of the measurement devices rely heavily on aspects like, preferences, variables, and availability and of course costs (Zikos and Petridis, 2014).

One more key element in the SLAM method is acquiring the data on the subject of the environmental background of the unmanned vehicle or robot. Similar to that of a human being, the unmanned vehicle or the robot uses the landmarks in determining its locations by the use of sensors attached to it. Regardless of the device that is being used for the purpose the medium i.e. the unmanned vehicle or robot is capable of observing the data and compute it on a constant basis. The robot or the unmanned vehicle will be using different landmarks for the various environments it operates in. however, there are certain dispositions that are required to be kept in mind the technique has not developed to the fullest possible extent thus, the robot or the unmanned vehicle will only be capable of determining landmarks which are stationary. As already stated the technique and the algorithms have not developed to the fullest possible extent and so it will not be able to assess a landmark that is moving constantly. In addition, the landmarks in order to be estimated are required to be distinguishable and can be recognized from the regular environment. Lastly the landmarks should be dense and have the possibility of viewing it from a wide range of angles.

Nevertheless, in recent researches it has been observed that SLAM with DATMO Detection and Tracking of Moving Objects is not only capable of operating in dynamic environments but also is fully capable of tracking and detailing the objects consisting the dynamic environment. The basis of the operating algorithm has been derived from the Bayesian formula, which provides the best possible apprehension on this particular subject matter.

The SLAM technique deserves appraisal and it gained appraisal from the international robotics community, international technological gazettes and other communities concentrated on the science and technology field. It was the wide scope the technology had bestowed upon mankind that it deserved appreciation. For instance, this technique could be implemented in exploring and mapping the interior of a human or animal body. This was not possible and is not possible in the present times without injuring the subject’s body. However, with the advancement of science and technology in the years to come this could used in exploring and mapping the interiors of organic bodies like the organisms consisting the biological world, including humans. For this purpose the technology needs to development minute robots or AI bots that would be minute enough to be inserted in the body without physically harming the body. Once inside the body the bots would be able to move inside the body constantly observing and computing the landmarks and it would be able compute its own location as well. This would provide the researchers with the knowledge of the points that the bots have explored and the current location of the bots. Moreover, as the Mars Rover is maneuvered on the surface of Mars, the researchers carrying out the instanced project would be able to maneuver the bots.

Apart from these aspects that SLAM has achieved as mapping and localization estimating method, there are persisting problems in the computational complexity and data association that need to be resolved fully and comprehensively. For instance identifying multiple and confusing landmarks still persists to be a challenge for the researchers. However, a recent research acquired a significant breakthrough in the SLAM literature based on the feature of re-evaluation of probabilistic foundation for SLAM i.e. Simultaneous Localization and Mapping. In this breakthrough it demonstrated its efficiency in terms of multiple objects Bayesian method of filtering. This filtering method incorporated finite sets aimed at providing advanced performance to leading SLAM algorithms featuring challenging measurements for scenarios having greater false alarm rates and greater missed detection rates by means of minimizing the requirement for data association. Two of the accepted techniques or filters for handling more than one object are Probability hypothesis density filter and the Joint Probabilistic Data Association.

Joint Probabilistic data association filter or JPDAF is actually an extension of the probabilistic data association filter, which is responsible for the assumption that one of the many potent entrants to the track is the correct one. It consists of an approach based on statistical analysis and deals with the issues relating to plot associations in radar tracker. In this process all the potential entrants posed for the association to a track are amalgamated in a sole and positive statistical probable update taking into account the precise statistical distribution of the clutters and errors in the track. This further requires assuming that one among the multiple entrants is the target and rest of the entrants is merely false alarms. It is one of the effective SLAM techniques used for the purpose of visual tracking in the computer vision field. 

The SLAM technique is undoubtedly having great potential in analyzing and mapping the sectors and environments which are out of the human reach. Therefore, the modern science and technology can benefit from this technology to great extent, if the innovators are prepared to build on it. The defense system of a country could be given new possibilities as this technique would offer them with the access to perform surveillances without attracting the attention of the enemies. In the earlier times surveillance systems incorporated photography and video capturing through aerial vehicles like jets and drones. In these scenarios the cameras were fixed to drones and the pilot had the access to remotely capture images from the cockpit. This led to many hassles and even loss of property and valuable lives. Therefore, incorporating advanced visual tracking systems available in the computer vision field would help in aiding such sort of surveillance systems.

Useful realizations of probabilistic SLAM have turned out to be more and more remarkable in the recent years. Capable of covering larger regions in more difficult environments have become a whole lot easier with the due passage of time. Here discussions on two implementations that will represent the major successes and mention other notable applications have been provided.

The “explore and return" experiment carried out by Newman was a moderate-scale indoor implementation SLAM technique that validated the non-divergent features of EKF-SLAM, which returned back to the specifically marked commencing position. The end result of the experimentation was extraordinary because the return trip was completely autonomous. The robot had been driven manually at the time of the exploration phase, although it did not have any visual contact with the operator. The operator solely relied on a real-time rendering of the map that the robot had been drawing up as a consequence of its exploration. During its return trip, the robot by means of the data that it had computed earlier had drawn its very own path and returned to the point from which it had started its journey without any sort of human intervention. This research demonstrated the capability of the concept and the results it might bring when used in the medical research field. It has already been discussed in the critical analysis section however; it was not fully discussed as to how the bots that will be inserted inside the human interiors would be able to function. Therefore, the research study was introduced in this part to provide an apprehension on the real time analysis on the way the technique will be functioning.

This technique if applied to the medical sciences would greatly benefit the people in the wider perspective. The technique would not only help the practitioners deduce the disorder but also they will be able to analyze the intense of the disorder. For instance, in case of a severe fracture if the technique is used the practitioners would be able to determine the extent of the fracture and take necessary steps for the treatment.

This technique would also be a great help in underwater exploration. Similar to the aerial extra-terrestrial exploration this kind of mapping and localization has posed a lot of challenges over the years. Since the submarines and underwater vehicles have limited accessibility as a result of their large size using small robots would help greatly in gaining access to those spots where the access cannot be gained by using other vehicles. 

Conclusion

The purpose of this essay was to present the SLAM technique involved in mapping and localizations purposes. This essay consisted of an introduction to the topic which particularly focused on elaborating the concept of SLAM. The introductory part dealt with the method by means of which SLAM technique operates and the purposes it is capable of serving. In the subsequent section an analysis is presented that deals with the brief history of the SLAM concept. It is in this section that an elaborate discussion is presented on the functioning of the SLAM technique and the factors that are directly related with its appropriate functioning. In the section subsequent this one critical analysis on the functioning of SLAM technique has been presented to offer an in-depth apprehension on the impact SLAM technology has made in the field of science and technology. This particular section of the essay also dealt with the purposes that the SLAM technique can serve with the advancement of Science and technology and lastly a few recommendations on the ways the technique can be implemented with the orientation to innovate and create better technological aspects. 

References

Ahn, S., Doh, N. and Chung, W. (2015). ROBUST NAVIGATION TECHNIQUES FOR THE GVG-BASED SLAM IN UNSTRUCTURED ENVIRONMENT. IFAC Proceedings Volumes, 38(1), pp.463-468.

Buck, M. (2014). Air Nike Slam Dunk. Feminist Studies, 30(2), p.271.

Carmack, H. (2014). Slam This: Understanding Language Choice and Delivery in Argument Using Slam Poetry. Communication Teacher, 23(1), pp.19-22.

Corbett, B. (n.d.). The big slam.

Crew, M., Phanavanh, B. and Garcia-Borges, C. (2014). Sequence and mRNA expression of nonclassical SLA class I genes SLA-7 and SLA-8. Immunogenetics, 56(2), pp.111-114.

Ferreira, F., Veruggio, G., Caccia, M. and Bruzzone, G. (2014). Speeded Up Robust Features for vision-based underwater motion estimation and SLAM: comparison with correlation-based techniques.IFAC Proceedings Volumes, 42(18), pp.273-278.

Gentile, C. (2013). Geolocation techniques. New York: Springer.

Havangi, R. (2015). Robust SLAM: SLAM base on $$hbox {H}_{infty }$$ H ∞ square root unscented Kalman filter. Nonlinear Dynamics, 83(1-2), pp.767-779.

He, B., Zhang, S., Yan, T., Zhang, T., Liang, Y. and Zhang, H. (2013). A Novel Combined SLAM Based on RBPF-SLAM and EIF-SLAM for Mobile System Sensing in a Large Scale Environment.Sensors, 11(12), pp.10197-10219.

Kim, J. (2012). Advanced methods, techniques, and applications in modeling and simulation. Tokyo: Springer.

Köster, V. (2014). First Science Slam in Karlsruhe. ChemViews.

Milstein, A. (2012). Improved particle filter based localization and mapping techniques. Waterloo, Ont.: University of Waterloo.

Oh, J. and Sim, K. (2014). Symmetrical model based SLAM : M-SLAM. Journal of Korean institute of intelligent systems, 20(4), pp.463-468.

Rethi, B. (2016). SLAM/SLAM interactions inhibit CD40-induced production of inflammatory cytokines in monocyte-derived dendritic cells. Blood, 107(7), pp.2821-2829.

Rich, J. (2015). SLAM. Upper Saddle River, N.J.: Pearson/Prentice Hall.

Schüz, P. (2012). „Sermon Slam“. Praktische Theologie, 46(1).

Sowter, T., Hale, G. and Starr, N. (2013). Bridge. New York, NY: Tess Press.

Szewczyk, R., Zieliński, C. and Kaliczyńska, M. (n.d.). Progress in automation, robotics and measuring techniques.

Taibi, T. (2014). Design patterns formalization techniques. Hershey, PA: IGI Pub.

Williams, B., Cummins, M., Neira, J., Newman, P., Reid, I. and Tardós, J. (2013). A comparison of loop closing techniques in monocular SLAM. Robotics and Autonomous Systems, 57(12), pp.1188-1197.

Williams, F. and Canedy, S. (2012). SLAM. Fort Monroe, Va.: Office of the Command Historian, U.S. Army Training and Doctrine Command.

YUAN, X. and ZHAO, C. (2011). 3D-SLAM Based on Point-plane Matching. ROBOT, 33(2), pp.215-221.

Zikos, N. and Petridis, V. (2014). 6-DoF Low Dimensionality SLAM (L-SLAM). Journal of Intelligent & Robotic Systems, 79(1), pp.55-72.