You are on page 1of 6

Research Proposal



Localization and Mapping is the most important and fundamental problem in developing Cyber
Phisical sytems like Robotics. The idea of simultaneously finding the pose of the device (Localization) and extracting the features of the unknown environment (Mapping), gave a breakthrough in
1995 and became an eyeopener in 2005, when Stanley autonomous car won DARPA grand challenge, which actually included Simultaneous Localization and Mapping (SLAM) system. The
solution to SLAM problem has been seen as a holy grail for the mobile robotics community as it
would provide the means to make a robot truly autonomous. SLAM algorithms are employed in unmanned aerial vehicles, autonomous underwater vehicles, planetary rovers, Industrial systems and
even inside the human body. Solving SLAM problem efficiently demands robust solutions to Data
Association, Feature extraction and Convergence related problems. Addressing these problems
to dynamically changing environments make the SLAM prolem more challenging and interesting.
Moreover best possible Data Fusion techniques are to be developed, as multi-robot deployment has
become a recent fashion to share computations and hence reducing time complexity. Interestingly
Optimization based approaches can provide more accurate solutions to SLAM problem and hence
making possible to develop interesting applications such as, Perpetual life assistants for older or
disabled people, Autonomous Vacuum Cleaner, The distributed autonomous gardening system,
Self-driving Cars and many more.



Robotic navigation, particularly when an external location reference such as a global positioning
system (GPS) is not available (eg: Indoor locations), requires the robot to be able to build a map
of the unknown environment in real-time and simultaneously work out its own location within the
map. Robust solutions to the Simultaneous Localization and Mapping (SLAM) problem, therefore,
underpins successful robot deployment in many application. The essential problem in SLAM is to
estimate robot location and the map of the environment, typically represented by a set of geometric
features, as measurements are gathered from a sensor as the robot moves through the environment.
The important point to note is, there is no single best solution to SLAM problem [1]. The method
chosen depends on number of factors.


Taxonomy of SLAM problem

Most important research papers identify the type of problems addressed by making the underlying
assumptions explicit in the following factors,
(i) Static vs Dyanmic : Static SLAM algorithms assume that the environment does not change
over time. Dynamic methods allow for changes in the environment. The vast majority of the
literature on SLAM assumes static environments. Mapping unstructured largesclae dynamic environments reamins an open research problem.

instead of the entire path. (v) Online vs Offline SLAM : The algorithms for the offline SLAM problem are often batch. . that is. ∗ mi : a vector describing the location of the ith landmark whose true location is assumed time invariant. although recently the problem of multirobot exploration [3] has gained in popularity. In the literature such algorithms are typically called filters. 3 Statement of SLAM problem and Related Issues Consider a mobile robot moving through an environment taking relative observations of a number of unknown landmarks using a sensor located on the robot. (iii) Known vs Unknown Correspondence : The correspondence problem is the problem of relating the identity of sensed things to other sensed things. the following quantities are defined as shown in figure 1 : Figure 1: simultaneous estimate of robot and landmarks Figure 2: Map and robot trajectory from SAM ∗ xk : the state vector describing the location and orientation of the vehicle.e correspondence is known) provide special mechanisms for estimating the correspondence of measured features to previously observed landmarks in the map. Metric SLAM methods provide metric information between the relation of such places. At a time instant k. Online SLAM seeks to recover the present robot location. they process all data at the same time [4]. (iv) Single-Robot Versus Multirobot SLAM : Most SLAM problems are defined for a singlerobot platform. The algorithms that do not make assumptions (i. applied at time k − 1 to drive the vehicle to a state xk at time k. The problem of estimating the correspondence is known as the data association problem [2]. ∗ zik : an observation taken from the vehicle of the location of the ith landmark at time k. ∗ uk : the control vector. Algorithms that address the online problem are usually incremental and can process one data item at a time. By fusing data collected by different robots we can gain more insights about the environment and moreover the job of computations can be divided among them so that computation time is reduced.(ii) Topological vs Metric : A topological map might be defined over a set of distinct places and a set of qualitative relations between these places.

. . we formulate the problem in probabilistic form that requires the probability distribution P (xk . x0 ) to be computed for all times k. x1 . m) + w(k) (3) where w(k) is the sensor observation noise at time k.e the Extended Kalman Filter (EKF) for representing the robots best estimate [5]. mn } : the set of all landmarks. perhaps through use of external information such as known landmarks remain interesting challenges. x0 ) P (zk |Z0:k−1 . x0 ) = P (zk |xk . xk } = {X0:k−1 . Virtually all SLAM literature assumes process models with additive noise of the form X (k + 1) = f (X(k). • Finally Particle Filter which applies nonparametric density estimation and efficient factorization methods to the SLAM problem [5]. m2 .In addition the following sets are also defined : ∗ X0:k = {x0 . uk } : the history of control inputs. . The three main paradigms from which a huge number of recently published methods are derived are • First comes the traditional approach. U0:k . . • Second one uses the fact that the SLAM can be viewed as a Sparse Graph of constraints. m|Z0:k . process noise and robot pose at time k repectively. U0:k . ∗ m = {m1 . x0 ) dxk−1 (1) Correction Update : P (xk . . Prediction Update : Z P (xk . m|Z0:k−1 . u2 . Since the control inputs and landmarks observations are prone to noise. . . . zk } : the set of all landmark observations. This can be done in a standard two-step recursive (sequential) prediction (timeupdate) correction (measurement-update) form as shown below. . and it applies nonlinear optimization [1] for recovering the map and the robots locations.1 Sensor and Process Models The sensor observation z(k) at time k is a function of robot pose X(k) and state of environment m is given by z (k) = h (X(k). . 3. m|Z0:k−1 . ∗ Z0:k = {z1 . U0:k . U0:k . u(k)) + v(k) (4) where u(k). . Theoretical problems posed when the noises are not zero-mean as well as practical solutions to this problem. . U0:k−1 . z2 . . uk } = {U0:k−1 . zk } = {Z0:k−1 . v(k). m|Z0:k−1 . U0:k ) (2) Solutions to the above probabilistic SLAM problem involve finding an appropriate representation for both the observation model and motion model that allows efficient and consistent computation of the prior and posterior distributions in (1) and (2). i. X(k) are the control. x0 ) = P (xk |xk−1 . m) P (xk . . uk ) × P (xk−1 . . m|Z0:k . . . ∗ U0:k = {u1 . xk } : the history of vehicle locations.

Both the EKF and particle filter based solutions to SLAM can produce inconsistent estimates. complex and large scale environments using vision as the sole external sensor i. Recent research demonstrated that use of Stochastic Gradient Descent. Most of the existing optimization based SLAM algorithms are based on Gauss-Newton or Levenberg-Marquardt optimizers that requires an initial guess to the robot poses and the map.A number of researchers conrm this fact for both the linear case [2] and the nonlinear case [8] in their work on EKF based SLAM algorithms.3. feature location uncertainty will monotonically decrease during SLAM. is also my focus of research. SLAM for dynamic. and energy efficient. 3. Finally I intend to work on improvements in computational efficiency by fromulating the SLAM problem as Nonnegative Linear Least Squares and employing Dimensionality Reduction[15] and try to answer questions such as how to obtain a map with given accuracy in minimum time. In recent years there has been a growing interest on the SLAM consistency issue among the research community. 4 Research Focus I intend to focus on Optimization based approaches for solving SLAM problem. Example of such approach is Smoothing and Mapping [9] which estimate the complete robot trajectory and the map as shown in figure 2. Although extended Information lter (EIF) based algorithms can overcome the computational issues to some extent [7] the issue of inconsistency remained unresolved. lightweight. Despite many advances by [6] limitations. on condition that noise covariance matrix is spherical. as these approaches can provide more accurate and consistent solutions. Tree based Network Optimizer algorithm makes SLAM problem to achieve good convergence results eventhough it starts from a poor initial guess [10]. 5 Applications • Autonomous Cars : Intel survey discovered that 44 percent of American respondents said they would like to live in a driverless city. This shows the demand in the market to build resilient . so as to address the problem of Data Association and develop Data Fusion techniques (for example we can extract depth information by fusing two images. makes it impractical to use EKF to solve large-scale SLAM problems [6].e cameras. as cameras have become popular sensors in the robotics community and take the advantage of being cheap. This vision based or visual SLAM [11] [12] uses concepts from Computer Vision and Machine Learning [13] and Optimization as well.3 Convergence On condition that observation model is avilable and the feature states are noise free and static. or how to maximize the coverage with a fixed time horizon and a required map quality. Acheiving convergence for dynamic environments is still a challenging task. together with its quadratic computational complexity associated with the presence of a dense covariance matrix of order (3 + 2n) × (3 + 2n) where n is the number of landmarks. In general SLAM problem is non-linear and non-convex with large search space.)[14].2 Consistency and Computational Issues A solution to a dynamic estimation problem is said to be consistent if the estimate is unbiased and the estimated covariance matrix matches the real mean square error.

FastSLAM: A factored solution to the simultaneous localization and mapping problem” Proc. 20-May-2008 . Distributed Multirobot Exploration and Mapping” Proceedings of the IEEE (Volume:94 . Oussama Khatib. that finds applications inside human body. B. Limketkai. Springer Handbook of Robotics.” Robotics and Automation. P. Issue: 7 )July 2006. Schulz. Konolige.. Stewart. etc can be build. M. Intell. . Perpetual life assistant for old or disbles as shown in figure 5.1611 pages. [4] M Montemerlo. S. K. Body SLAM [17] enhances the positioning accuracy of WCE. WA Ko. J. . AAAI Nat.W. • Wireless Capsule Endoscopy : SLAM also try to enhance Intel’s vision towards Health and Life Sciences. . • Home Automation : Methods from Robotics (SLAM) and Machine Learning can be used to increase household efficiency and extend device functionality.F.. D. as it takes the advantage of Data Fusion of image sequences captured by the WCEs embedded camera and the RF signal emitted by the capsule. . [2] Dissanayake. To detect precise position of the intestinal disease we need to localize the capsule in the unknown map of the intestine. Clark. . which is very much desired in safe navigation for autonomous cars. Issue: 3 )Jun 2001. B Wegbreit. B.M. Seattle.. Durrant-Whyte. D Koller. Artif. .Figure 3: Wireless Capsule Endoscopy Figure 4: Vacuum cleaner Figure 5: Walker Assistant autonomous cars. Advancements in Optimization based approaches solves SLAM problem more accurately.. For example WCE (see figure 3) offers painless investigation of the entire small intestine of human body. pp. . S Thrun. M. Many researchers proved SLAM is a good substitute where ever GPS is absent and more erroneous.. by building systems such as Body-SLAM.” Springer science and business media. Newman. H. D. IEEE Transactions on (Volume:17 . Conf. Csorba.Computers .G. [3] Fox. . References [1] Bruno Siciliano. A solution to the simultaneous localization and map building (SLAM) problem.593 -598 2002 . Interesting devices such as Automatic Vacuum cleaner [16] in figure 4.

Pages 149157. Teller Fast iterative alignment of pose graphs with poor initial estimates. Thrun. 2010 20th International Conference on 23-26 Aug. 13(3):108-117. [9] F. and Artificial Intelligence Lab. 2006.G. Body-SLAM. Kaess. [16] WooYeon Jeong. ICRA 2006. and D. Tim. Simultanouse localization and mapping (SLAM): Part II. 25(12):1181-1203. . Issue: 3 )June 2005. Fox. Probabilistic Robotics. EKF-SLAM and Machine Learning Techniques for Visual Robot Navigation. W. Issue: 1 )22 December 2014. Vision-based global localization and mapping for mobile robots. 2005. CV-SLAM: a new ceiling vision-based SLAM technique. [14] Hollinger G. IEEE Transactions on (Volume:31 .. IEEE Transactions on 28 January 2011. G. MIT.” Robotics and Automation Magazine. Durrant-Whyte. Square root SAM: Simultaneous localization and mapping via square root information smoothing.” IEEE Transactions on Robotics. Huang.Robotics and Automation. Cambridge). 2005. [13] Casarrubias-Vargas H. January 2015.Robotics.Yerramalli S. 2005. IEEE Transactions on (Volume:21 .A. Petrilli-Barcelo A. [7] S. MA Leonard J.Intelligent Robots and Systems.. Bailey. Bailey and H. 2005 IEEE/RSJ International Conference on 2-6 Aug.[5] Durrant-Whyte H. Shoudong Huang. Sci. Lamp. and Door Features From Upward-Looking .Industrial Electronics. . [17] Guanqun Bao. [12] Seo-Yeon Hwang. 2010. 23(5):1036-1049. Jae-Bok Song Monocular Vision-Based SLAM in Indoor Environment Using Corner. 091313/unrestricted/main. Convergence and consistency analysis for Extended Kalman Filter Based SLAM.Pattern Recognition (ICPR).Robotics. [6] T.wpi. [10] Olson. Distributed Data Fusion for Multirobot Search. 2007.” IEEE Robotics and Automation Magazine. [8] S. https://www. IEEE (Volume:13 .” The MIT Press. E. Simultaneous localization and mapping: part I. (Comput.Automatica Volume 51. Dissanayake. Little. Issue: 2 )June 2006. (IROS 2005). International Journal of Robotics Research. Bayro-Corrochano E. Dellaert and M. 2006. [15] Heng Wanga. Dimensionality reduction for point feature SLAM problems with spherical covariance matrices. Lowe D. Kyoung Mu Lee. 2006.pdf. Burgard. [11] Se S. .