Skip to main content

Zoltán Vámossy

Óbuda University, John von Neumann Faculty of Informatics, Faculty Member

Followers

37

Following

14

Co-authors

9

Public Views

Associate professor

less

Harvard University

David Pierre Leibovitz

Carleton University

Carnegie Mellon University

Kansas State University

University of Cologne

Armando Marques-Guedes

UNL - New University of Lisbon

Oxford Brookes University

University of Cape Town

marcello frixione

University of Genova

Fusun Balik Sanli

Yildiz Technical University

InterestsView All (11)

Uploads

Papers by Zoltán Vámossy

Improved RGB-D Camera-based SLAM System for Mobil Robots

Acta polytechnica Hungarica, 2024

The paper presents an improved Simultaneous Localization and Mapping (iSLAM) and 3D reconstructio... more The paper presents an improved Simultaneous Localization and Mapping (iSLAM) and 3D reconstruction system for a mobile robot carrying an RGB-D camera. The developed system generates a three-dimensional point cloud from the color and depth camera data of the RGB-D camera. The matching is performed on successive point clouds that partially overlap. After feature detection on the color camera images, the method selects the 3D points in the successive point clouds that belong together. During the iterative multi-step matching algorithm based on Singular Value Decomposition (ISVD), it minimizes the matching error between the selected point clouds and deletes non-real point pairs in the process. Our previous SLAM method has been improved in several ways. On the one hand, a conditional averaging filter (Distance Image Filter: DIF) was created for the depth camera data to reduce the noise. The matching algorithm iteratively determines the matching transformation and the estimated displacement from the feature points of several recent point cloud segments. It defines a parameter for the accuracy and quality of each matching and includes the sub-results of pairs in the final displacement estimate by weighting these accuracy parameters. In this way, the algorithm yields significantly improved accuracy values, which in all cases are of comparable magnitude to those of methods in the literature, and for some published test sets exceed their characteristics. Since parallel programming methods are used to run the fits to previous states, the operation runtime remains fast. If the robot returns to its previous location, the improved loop closure detection method detects this fact, refines the estimates, and improves the accuracy. Finally, the proposed system also produces a 3D point cloud of the environment.

2サンプルT試験ベースの障害物認識を用いたロボットナビゲーションにおける障害物回避【JST・京大機械翻訳】

Lecture notes in electrical engineering, 2021

Benefits of Layered Software Architecture in Machine Learning Applications

Proceedings of the International Conference on Image Processing and Vision Engineering

Obstacle Avoidance during Robot Navigation in Dynamic Environment using Fuzzy Controller

International journal of recent technology and engineering, Jul 30, 2019

A Simulink model containing fuzzy logic controller for collision-free robot navigation in a dynam... more A Simulink model containing fuzzy logic controller for collision-free robot navigation in a dynamic environment is presented in this paper. Two controllers, pure pursuit and fuzzy logic controller, are considered to handle robot navigation with obstacle avoidance. Ignoring the obstacles, the pure pursuit controller computes the required linear and angular velocities to direct robot from start to goal location. However, if obstacles are present in the navigation path then the robot will get collided with obstacles in the path. As a result, the robot will not reach to the provided goal location. The fuzzy logic controller is used to avoid obstacles in the navigation path. The fuzzy logic controller takes obstacle distance, obstacle angle, target direction and the x coordinate of goal location as inputs. Consequently, the fuzzy logic controller outputs the required change in angular velocity for the robot. This change in angular velocity is applied to the angular velocity provided by the pure pursuit controller. The experimental work is performed using Turtlebot Gazebo simulator. The navigation including environment, obstacles and resultant paths are also manifested.

Navigation and Cognitive Techniques for Humanoid Robots

Solving Jigsaw Puzzles Using Computer Vision and Curve Similarity Measures

Image-based Lesion Classification using Deep Neural Networks

Proceedings of the 2nd International Conference on Image Processing and Vision Engineering

Proceedings of the 2nd International Conference on Image Processing and Vision Engineering

Occlusion handling is one of the challenges of object detection and segmentation, and scene under... more Occlusion handling is one of the challenges of object detection and segmentation, and scene understanding. Because objects appear differently when they are occluded in varying degree, angle, and locations. Therefore, determining the existence of occlusion between objects and their order in a scene is a fundamental requirement for semantic understanding. Existing works mostly use deep learning based models to retrieve the order of the instances in an image or for occlusion detection. This requires labelled occluded data and it is timeconsuming. In this paper, we propose a simpler and faster method that can perform both operations without any training and only requires the modal segmentation masks. For occlusion detection, instead of scanning the two objects entirely, we only focus on the intersected area between their bounding boxes. Similarly, we use the segmentation mask inside the same area to recover the depth-ordering. When tested on COCOA dataset, our method achieves +8% and +5% more accuracy than the baselines in order recovery and occlusion detection respectively.

Evaluation of colony formation dataset of simulated cell cultures

Applied Medical Informaticvs, 2020

In vitro biological experiments and in silico individual-based computational models are widely us... more In vitro biological experiments and in silico individual-based computational models are widely used to understand the low-level behavior of cells and cellular functions. Many of these functions can not be directly observed, however, may be deduced from other properties that can be well measured and modeled. In this paper, we present a procedure to evaluate synthetic cell colony formation generated by an off-lattice individual-based model. The calculated shape features of the artificial cell aggregates can be related to the parameter values of the simulated agents, therefore this data can be used to quantify properties of real-life cells such as motility or binding affinity that can not be easily determined otherwise. Our experiments showed that only a few of these parameters are responsible for the difference in shape features of the colonies.

Generative Adversarial Network for Overcoming Occlusion in Images: A Survey

Algorithms

Although current computer vision systems are closer to the human intelligence when it comes to co... more Although current computer vision systems are closer to the human intelligence when it comes to comprehending the visible world than previously, their performance is hindered when objects are partially occluded. Since we live in a dynamic and complex environment, we encounter more occluded objects than fully visible ones. Therefore, instilling the capability of amodal perception into those vision systems is crucial. However, overcoming occlusion is difficult and comes with its own challenges. The generative adversarial network (GAN), on the other hand, is renowned for its generative power in producing data from a random noise distribution that approaches the samples that come from real data distributions. In this survey, we outline the existing works wherein GAN is utilized in addressing the challenges of overcoming occlusion, namely amodal segmentation, amodal content completion, order recovery, and acquiring training data. We provide a summary of the type of GAN, loss function, the...

A novel method for Convolutional Neural Architecture Generation with memory limitation

2019 IEEE 17th World Symposium on Applied Machine Intelligence and Informatics (SAMI), 2019

A novel algorithm-based method for Neural Architecture Generation is introduced in this paper. Un... more A novel algorithm-based method for Neural Architecture Generation is introduced in this paper. Unlike NASNet or AutoML, which outputs one optimal trained model based on a Recurrent Neural Network, the presented method outputs multiple possible architectures, without training. This method is not meant for automatic model generation based on inputs, it is designed for analyzing input preprocessing methods. The classical convolutional design patterns are analyzed and the properties and validation steps for evaluation of the generated architectures are defined, including the estimated size of the model based on the parameter numbers and training batch sizes.

Vehicle Image Matching Using Siamese Neural Networks with Multi-Directional Image Projections

2018 IEEE 12th International Symposium on Applied Computational Intelligence and Informatics (SACI)

A novel method to measure similarity on images is based on the Siamese architecture of Convolutio... more A novel method to measure similarity on images is based on the Siamese architecture of Convolutional Neural Networks. Two identical CNNs extract the significant parts of the inputs, and a single neuron outputs the distance of the two. The aim of this paper is to introduce that the Siamese structure can be applied to match image projection based signatures of vehicles.

Multi-Directional Projection Transformations for Machine Learning based Object Matching

2019 IEEE 13th International Symposium on Applied Computational Intelligence and Informatics (SACI)

In case of object matching on low-quality images, projection-based similarity measurement is an o... more In case of object matching on low-quality images, projection-based similarity measurement is an often used approach. In recent advances, deep convolutional neural networks are used to extract the features from the input data. In this paper, an analysis is done to compare end-to-end methods with projection-based preprocessing methods, including the Radon transform and the Trace transform. The paper explains the method of convolutional neural architecture generation based on input sizes, followed by the parallel processing on a cluster of workstations. To show the applicability of multi-directional projections as a preprocessing method on small storage units, model memory cost and prediction performance are both measured and optimal results are selected accordingly.

BBBD: Bounding Box Based Detector for Occlusion Detection and Order Recovery

Proceedings of the 2nd International Conference on Image Processing and Vision Engineering

Occlusion handling is one of the challenges of object detection and segmentation, and scene under... more Occlusion handling is one of the challenges of object detection and segmentation, and scene understanding. Because objects appear differently when they are occluded in varying degree, angle, and locations. Therefore, determining the existence of occlusion between objects and their order in a scene is a fundamental requirement for semantic understanding. Existing works mostly use deep learning based models to retrieve the order of the instances in an image or for occlusion detection. This requires labelled occluded data and it is timeconsuming. In this paper, we propose a simpler and faster method that can perform both operations without any training and only requires the modal segmentation masks. For occlusion detection, instead of scanning the two objects entirely, we only focus on the intersected area between their bounding boxes. Similarly, we use the segmentation mask inside the same area to recover the depth-ordering. When tested on COCOA dataset, our method achieves +8% and +5% more accuracy than the baselines in order recovery and occlusion detection respectively.

ISVD-Based Advanced Simultaneous Localization and Mapping (SLAM) Algorithm for Mobile Robots

Machines

In the case of simultaneous localization and mapping, route planning and navigation are based on ... more In the case of simultaneous localization and mapping, route planning and navigation are based on data captured by multiple sensors, including built-in cameras. Nowadays, mobile devices frequently have more than one camera with overlapping fields of view, leading to solutions where depth information can also be gathered along with ordinary RGB color data. Using these RGB-D sensors, two- and three-dimensional point clouds can be recorded from the mobile devices, which provide additional information for localization and mapping. The method of matching point clouds during the movement of the device is essential: reducing noise while having an acceptable processing time is crucial for a real-life application. In this paper, we present a novel ISVD-based method for displacement estimation, using key points detected by SURF and ORB feature detectors. The ISVD algorithm is a fitting procedure based on SVD resolution, which removes outliers from the point clouds to be fitted in several steps...

Multiprocessing of an individual-cell based model for parameter testing

2016 IEEE 11th International Symposium on Applied Computational Intelligence and Informatics (SACI), 2016

To simulate the behaviour of cell-level biological interactions, cells can be modeled as individu... more To simulate the behaviour of cell-level biological interactions, cells can be modeled as individual agents interacting with each other. To identify the effects of parameter perturbation, a massive number of simulations are usually required that are feasible to carry out by parallel computation. In this article, a multiprocessing scheme is defined to reduce the overall runtime. The scheme relies on predicting the expected computational cost of each input parameter set, and scheduling them based on the Longest Processing Time (LPT) heuristic rule.

Application and properties of the radon transform for object image matching

2017 IEEE 15th International Symposium on Applied Machine Intelligence and Informatics (SAMI), 2017

The Radon transform and the reverse formulas are mostly used in computer tomography to reconstruc... more The Radon transform and the reverse formulas are mostly used in computer tomography to reconstruct the original shape. The projections of an image could be used in computer vision for object recognition, based on projection matching. By comparing the projections, an effective method could be defined. To understand the projection behavior the basic properties of the transform are inspected, and the results of several image transformations are analyzed.

Densely Connected Convolutional Networks (DenseNet) for Diagnosing Coronavirus Disease (COVID-19) from Chest X-ray Imaging

2021 IEEE International Symposium on Medical Measurements and Applications (MeMeA), 2021

Since the beginning of the coronavirus disease (COVID-19) pandemic several machine learning and d... more Since the beginning of the coronavirus disease (COVID-19) pandemic several machine learning and deep learning methods had been introduced to detect the infected patients using the X-Ray or CT scan images. Numerous sophisticated data-driven methods had been introduced to improve the performance and the accuracy of the diagnosis models. This paper proposes an improved densely connected convolutional networks (DenseNet) method based on transfer learning (TL) to enhance the model performance. The results show promising model accuracy.

Laser Scan Matching in Robot Navigation

2018 IEEE 12th International Symposium on Applied Computational Intelligence and Informatics (SACI), 2018

In this study, Laser scan matching is implemented for simultaneous localization and mapping in ro... more In this study, Laser scan matching is implemented for simultaneous localization and mapping in robot navigation. Two algorithms are presented to accomplish the laser scan matching using normal distribution transform. Firstly, an algorithm is proposed to collect and store the laser scan data obtained from the robot scan. The second algorithm is to perform scan matching and map building. Using the data of good scan matches, a neural network is trained and tested. The experimental work is carried out on the Turtlebot robot model in Gazebo simulator using MATLAB software.

Robot navigation in unknown environment using fuzzy logic

2017 IEEE 15th International Symposium on Applied Machine Intelligence and Informatics (SAMI), 2017

In this paper, a Simulink model for the robot navigation in unknown environment is presented. The... more In this paper, a Simulink model for the robot navigation in unknown environment is presented. The robot navigation is handled by two controllers: pure pursuit and fuzzy logic controller. The pure pursuit controller computes a direct path from start to goal position without considering the obstacles in the path. For obstacle avoidance in robot navigation, the fuzzy logic controller is taken. This fuzzy logic controller takes the input from the laser sensor of the robot and gives the change in the angular velocity as output to the robot to avoid the obstacle. The navigation paths resulting from the proposed Simulink model, with and without obstacles in the paths, are shown in figures.

Improved RGB-D Camera-based SLAM System for Mobil Robots

Acta polytechnica Hungarica, 2024

The paper presents an improved Simultaneous Localization and Mapping (iSLAM) and 3D reconstructio... more The paper presents an improved Simultaneous Localization and Mapping (iSLAM) and 3D reconstruction system for a mobile robot carrying an RGB-D camera. The developed system generates a three-dimensional point cloud from the color and depth camera data of the RGB-D camera. The matching is performed on successive point clouds that partially overlap. After feature detection on the color camera images, the method selects the 3D points in the successive point clouds that belong together. During the iterative multi-step matching algorithm based on Singular Value Decomposition (ISVD), it minimizes the matching error between the selected point clouds and deletes non-real point pairs in the process. Our previous SLAM method has been improved in several ways. On the one hand, a conditional averaging filter (Distance Image Filter: DIF) was created for the depth camera data to reduce the noise. The matching algorithm iteratively determines the matching transformation and the estimated displacement from the feature points of several recent point cloud segments. It defines a parameter for the accuracy and quality of each matching and includes the sub-results of pairs in the final displacement estimate by weighting these accuracy parameters. In this way, the algorithm yields significantly improved accuracy values, which in all cases are of comparable magnitude to those of methods in the literature, and for some published test sets exceed their characteristics. Since parallel programming methods are used to run the fits to previous states, the operation runtime remains fast. If the robot returns to its previous location, the improved loop closure detection method detects this fact, refines the estimates, and improves the accuracy. Finally, the proposed system also produces a 3D point cloud of the environment.

2サンプルT試験ベースの障害物認識を用いたロボットナビゲーションにおける障害物回避【JST・京大機械翻訳】

Lecture notes in electrical engineering, 2021

Benefits of Layered Software Architecture in Machine Learning Applications

Proceedings of the International Conference on Image Processing and Vision Engineering

Obstacle Avoidance during Robot Navigation in Dynamic Environment using Fuzzy Controller

International journal of recent technology and engineering, Jul 30, 2019

A Simulink model containing fuzzy logic controller for collision-free robot navigation in a dynam... more A Simulink model containing fuzzy logic controller for collision-free robot navigation in a dynamic environment is presented in this paper. Two controllers, pure pursuit and fuzzy logic controller, are considered to handle robot navigation with obstacle avoidance. Ignoring the obstacles, the pure pursuit controller computes the required linear and angular velocities to direct robot from start to goal location. However, if obstacles are present in the navigation path then the robot will get collided with obstacles in the path. As a result, the robot will not reach to the provided goal location. The fuzzy logic controller is used to avoid obstacles in the navigation path. The fuzzy logic controller takes obstacle distance, obstacle angle, target direction and the x coordinate of goal location as inputs. Consequently, the fuzzy logic controller outputs the required change in angular velocity for the robot. This change in angular velocity is applied to the angular velocity provided by the pure pursuit controller. The experimental work is performed using Turtlebot Gazebo simulator. The navigation including environment, obstacles and resultant paths are also manifested.

Navigation and Cognitive Techniques for Humanoid Robots

Solving Jigsaw Puzzles Using Computer Vision and Curve Similarity Measures

Image-based Lesion Classification using Deep Neural Networks

Proceedings of the 2nd International Conference on Image Processing and Vision Engineering

Proceedings of the 2nd International Conference on Image Processing and Vision Engineering

Occlusion handling is one of the challenges of object detection and segmentation, and scene under... more Occlusion handling is one of the challenges of object detection and segmentation, and scene understanding. Because objects appear differently when they are occluded in varying degree, angle, and locations. Therefore, determining the existence of occlusion between objects and their order in a scene is a fundamental requirement for semantic understanding. Existing works mostly use deep learning based models to retrieve the order of the instances in an image or for occlusion detection. This requires labelled occluded data and it is timeconsuming. In this paper, we propose a simpler and faster method that can perform both operations without any training and only requires the modal segmentation masks. For occlusion detection, instead of scanning the two objects entirely, we only focus on the intersected area between their bounding boxes. Similarly, we use the segmentation mask inside the same area to recover the depth-ordering. When tested on COCOA dataset, our method achieves +8% and +5% more accuracy than the baselines in order recovery and occlusion detection respectively.

Evaluation of colony formation dataset of simulated cell cultures

Applied Medical Informaticvs, 2020

In vitro biological experiments and in silico individual-based computational models are widely us... more In vitro biological experiments and in silico individual-based computational models are widely used to understand the low-level behavior of cells and cellular functions. Many of these functions can not be directly observed, however, may be deduced from other properties that can be well measured and modeled. In this paper, we present a procedure to evaluate synthetic cell colony formation generated by an off-lattice individual-based model. The calculated shape features of the artificial cell aggregates can be related to the parameter values of the simulated agents, therefore this data can be used to quantify properties of real-life cells such as motility or binding affinity that can not be easily determined otherwise. Our experiments showed that only a few of these parameters are responsible for the difference in shape features of the colonies.

Generative Adversarial Network for Overcoming Occlusion in Images: A Survey

Algorithms

Although current computer vision systems are closer to the human intelligence when it comes to co... more Although current computer vision systems are closer to the human intelligence when it comes to comprehending the visible world than previously, their performance is hindered when objects are partially occluded. Since we live in a dynamic and complex environment, we encounter more occluded objects than fully visible ones. Therefore, instilling the capability of amodal perception into those vision systems is crucial. However, overcoming occlusion is difficult and comes with its own challenges. The generative adversarial network (GAN), on the other hand, is renowned for its generative power in producing data from a random noise distribution that approaches the samples that come from real data distributions. In this survey, we outline the existing works wherein GAN is utilized in addressing the challenges of overcoming occlusion, namely amodal segmentation, amodal content completion, order recovery, and acquiring training data. We provide a summary of the type of GAN, loss function, the...

A novel method for Convolutional Neural Architecture Generation with memory limitation

2019 IEEE 17th World Symposium on Applied Machine Intelligence and Informatics (SAMI), 2019

A novel algorithm-based method for Neural Architecture Generation is introduced in this paper. Un... more A novel algorithm-based method for Neural Architecture Generation is introduced in this paper. Unlike NASNet or AutoML, which outputs one optimal trained model based on a Recurrent Neural Network, the presented method outputs multiple possible architectures, without training. This method is not meant for automatic model generation based on inputs, it is designed for analyzing input preprocessing methods. The classical convolutional design patterns are analyzed and the properties and validation steps for evaluation of the generated architectures are defined, including the estimated size of the model based on the parameter numbers and training batch sizes.

Vehicle Image Matching Using Siamese Neural Networks with Multi-Directional Image Projections

2018 IEEE 12th International Symposium on Applied Computational Intelligence and Informatics (SACI)

A novel method to measure similarity on images is based on the Siamese architecture of Convolutio... more A novel method to measure similarity on images is based on the Siamese architecture of Convolutional Neural Networks. Two identical CNNs extract the significant parts of the inputs, and a single neuron outputs the distance of the two. The aim of this paper is to introduce that the Siamese structure can be applied to match image projection based signatures of vehicles.

Multi-Directional Projection Transformations for Machine Learning based Object Matching

2019 IEEE 13th International Symposium on Applied Computational Intelligence and Informatics (SACI)

In case of object matching on low-quality images, projection-based similarity measurement is an o... more In case of object matching on low-quality images, projection-based similarity measurement is an often used approach. In recent advances, deep convolutional neural networks are used to extract the features from the input data. In this paper, an analysis is done to compare end-to-end methods with projection-based preprocessing methods, including the Radon transform and the Trace transform. The paper explains the method of convolutional neural architecture generation based on input sizes, followed by the parallel processing on a cluster of workstations. To show the applicability of multi-directional projections as a preprocessing method on small storage units, model memory cost and prediction performance are both measured and optimal results are selected accordingly.

BBBD: Bounding Box Based Detector for Occlusion Detection and Order Recovery

Proceedings of the 2nd International Conference on Image Processing and Vision Engineering

Occlusion handling is one of the challenges of object detection and segmentation, and scene under... more Occlusion handling is one of the challenges of object detection and segmentation, and scene understanding. Because objects appear differently when they are occluded in varying degree, angle, and locations. Therefore, determining the existence of occlusion between objects and their order in a scene is a fundamental requirement for semantic understanding. Existing works mostly use deep learning based models to retrieve the order of the instances in an image or for occlusion detection. This requires labelled occluded data and it is timeconsuming. In this paper, we propose a simpler and faster method that can perform both operations without any training and only requires the modal segmentation masks. For occlusion detection, instead of scanning the two objects entirely, we only focus on the intersected area between their bounding boxes. Similarly, we use the segmentation mask inside the same area to recover the depth-ordering. When tested on COCOA dataset, our method achieves +8% and +5% more accuracy than the baselines in order recovery and occlusion detection respectively.

ISVD-Based Advanced Simultaneous Localization and Mapping (SLAM) Algorithm for Mobile Robots

Machines

In the case of simultaneous localization and mapping, route planning and navigation are based on ... more In the case of simultaneous localization and mapping, route planning and navigation are based on data captured by multiple sensors, including built-in cameras. Nowadays, mobile devices frequently have more than one camera with overlapping fields of view, leading to solutions where depth information can also be gathered along with ordinary RGB color data. Using these RGB-D sensors, two- and three-dimensional point clouds can be recorded from the mobile devices, which provide additional information for localization and mapping. The method of matching point clouds during the movement of the device is essential: reducing noise while having an acceptable processing time is crucial for a real-life application. In this paper, we present a novel ISVD-based method for displacement estimation, using key points detected by SURF and ORB feature detectors. The ISVD algorithm is a fitting procedure based on SVD resolution, which removes outliers from the point clouds to be fitted in several steps...

Multiprocessing of an individual-cell based model for parameter testing

2016 IEEE 11th International Symposium on Applied Computational Intelligence and Informatics (SACI), 2016

To simulate the behaviour of cell-level biological interactions, cells can be modeled as individu... more To simulate the behaviour of cell-level biological interactions, cells can be modeled as individual agents interacting with each other. To identify the effects of parameter perturbation, a massive number of simulations are usually required that are feasible to carry out by parallel computation. In this article, a multiprocessing scheme is defined to reduce the overall runtime. The scheme relies on predicting the expected computational cost of each input parameter set, and scheduling them based on the Longest Processing Time (LPT) heuristic rule.

Application and properties of the radon transform for object image matching

2017 IEEE 15th International Symposium on Applied Machine Intelligence and Informatics (SAMI), 2017

The Radon transform and the reverse formulas are mostly used in computer tomography to reconstruc... more The Radon transform and the reverse formulas are mostly used in computer tomography to reconstruct the original shape. The projections of an image could be used in computer vision for object recognition, based on projection matching. By comparing the projections, an effective method could be defined. To understand the projection behavior the basic properties of the transform are inspected, and the results of several image transformations are analyzed.

Densely Connected Convolutional Networks (DenseNet) for Diagnosing Coronavirus Disease (COVID-19) from Chest X-ray Imaging

2021 IEEE International Symposium on Medical Measurements and Applications (MeMeA), 2021

Since the beginning of the coronavirus disease (COVID-19) pandemic several machine learning and d... more Since the beginning of the coronavirus disease (COVID-19) pandemic several machine learning and deep learning methods had been introduced to detect the infected patients using the X-Ray or CT scan images. Numerous sophisticated data-driven methods had been introduced to improve the performance and the accuracy of the diagnosis models. This paper proposes an improved densely connected convolutional networks (DenseNet) method based on transfer learning (TL) to enhance the model performance. The results show promising model accuracy.

Laser Scan Matching in Robot Navigation

2018 IEEE 12th International Symposium on Applied Computational Intelligence and Informatics (SACI), 2018

In this study, Laser scan matching is implemented for simultaneous localization and mapping in ro... more In this study, Laser scan matching is implemented for simultaneous localization and mapping in robot navigation. Two algorithms are presented to accomplish the laser scan matching using normal distribution transform. Firstly, an algorithm is proposed to collect and store the laser scan data obtained from the robot scan. The second algorithm is to perform scan matching and map building. Using the data of good scan matches, a neural network is trained and tested. The experimental work is carried out on the Turtlebot robot model in Gazebo simulator using MATLAB software.

Robot navigation in unknown environment using fuzzy logic

2017 IEEE 15th International Symposium on Applied Machine Intelligence and Informatics (SAMI), 2017

In this paper, a Simulink model for the robot navigation in unknown environment is presented. The... more In this paper, a Simulink model for the robot navigation in unknown environment is presented. The robot navigation is handled by two controllers: pure pursuit and fuzzy logic controller. The pure pursuit controller computes a direct path from start to goal position without considering the obstacles in the path. For obstacle avoidance in robot navigation, the fuzzy logic controller is taken. This fuzzy logic controller takes the input from the laser sensor of the robot and gives the change in the angular velocity as output to the robot to avoid the obstacle. The navigation paths resulting from the proposed Simulink model, with and without obstacles in the paths, are shown in figures.