Large Scale SLAM in an Urban Environment

The simultaneous localisation and mapping (SLAM) problem is one of the most central in robotics research. It asks if a mobile robot, put in an unknown location in an unknown environment, can incrementally build a consistent map of the environment and simultaneously determine its location within this map. In this thesis, an attempt to solve the SLAM problem in constant time in a complex environment, such as a suburban area, is made. Such a solution must handle increasing amounts of data without significant increase in computation time.A delayed state information filter is used to estimate the robot’s trajectory, …


1 Introduction
1.1 Problem Formulation
1.2 Approach
1.3 Related Work
1.4 ACFR
1.5 Chapter Outline
2 Dynamic Motion Model
2.1 State Vector
2.1.1 State Vector in Information Form
2.1.2 Relationship Between Covariance and Information
2.2 Vehicle Motion Model
2.2.1 Steering Model
2.2.2 Turn Rate Model
2.2.3 Laser Model
2.2.4 Process Noise
3 Laser Signal Processing
3.1 The Laser Sensor
3.1.1 Laser Scans
3.2 Laser Matching
3.2.1 Features
3.2.2 Classifiers
3.2.3 Boosting
3.2.4 Training with Tenfold Cross Validation
3.3 Laser Scan Alignment
3.3.1 Iterative Closest Point
3.3.2 Conditional Random Field Matching
3.3.3 Covariance of Laser Scan Alignment
4 Image Processing
4.1 Feature Extraction
4.1.1 Region Detectors
4.1.2 Region Descriptors
4.1.3 Implemented Extractors
4.2 Tree of Words
4.2.1 Background
4.2.2 Building a Vocabulary Tree
4.2.3 Descriptor Classification
4.2.4 Add Image to Base
4.2.5 Compare Image to Base
4.2.6 Spatial Consistency
5.1 Exactly Sparse Delayed State Filter
5.1.1 Prediction with Augmentation
5.1.2 Prediction without Augmentation
5.1.3 Observation
5.1.4 Measurement Update
5.1.5 Recovering the State Vector
5.2 Candidates for Loop Closure
5.2.1 Initial Set of Loop Closure Candidates
5.2.2 Final Set of Loop Closure Candidates
5.2.3 Recovering the Covariance Matrix
5.3 Laser Scan Alignment
5.3.1 Laser Scan Alignment Tests
5.4 Innovation Gating
5.5 SLAM algorithm
5.6 Classification Definitions
5.7 ESDF simulations
6 Experimental Results: Image Based SLAM
6.1 Loop Closure Detection
6.1.1 Tree of Words in SLAM
6.1.2 Loop Closure Experiment Settings
6.1.3 Loop Closure Performance
6.2 Ground Vehicle Experiments
6.2.1 Data Set 1
6.2.2 Data Set 2
6.3 Underwater Experiments
6.4 Discussion
6.5 Future Work
7 Experimental Results: Laser Based SLAM
7.1 SLAM algorithm
7.2 Receiver Operating Characteristic
7.3 Test of Laser Scan Matching
7.4 Test of Rotation Invariance
7.5 Ground Vehicle Experiments
7.5.1 Data set 1
7.5.2 Data set 2
7.5.3 Data set 3
7.6 Discussion
7.7 Future Work
8 Conclusions
A K Means Clustering
B The Compounding Operations
B.1 Compounding
B.2 Inverse Relationship
B.3 Composite Relationships
B.4 Robot Examples
C Abbreviations

Author: Granström, Karl,Callmer, Jonas

Source: Linköping University

Download URL 2: Visit Now

Leave a Comment