Research Output
1 - 10 of 14
-
PublicationPerformance analysis of the microsoft kinect sensor for 2D Simultaneous Localization and Mapping (SLAM) techniques( 2014)This paper presents a performance analysis of two open-source, laser scanner-based Simultaneous Localization and Mapping (SLAM) techniques (i.e., Gmapping and Hector SLAM) using a Microsoft Kinect to replace the laser sensor. Furthermore, the paper proposes a new system integration approach whereby a Linux virtual machine is used to run the open source SLAM algorithms. The experiments were conducted in two different environments; a small room with no features and a typical office corridor with desks and chairs. Using the data logged from real-time experiments, each SLAM technique was simulated and tested with different parameter settings. The results show that the system is able to achieve real time SLAM operation. The system implementation offers a simple and reliable way to compare the performance of Windows-based SLAM algorithm with the algorithms typically implemented in a Robot Operating System (ROS). The results also indicate that certain modifications to the default laser scanner-based parameters are able to improve the map accuracy. However, the limited field of view and range of Kinect's depth sensor often causes the map to be inaccurate, especially in featureless areas, therefore the Kinect sensor is not a direct replacement for a laser scanner, but rather offers a feasible alternative for 2D SLAM tasks.
-
PublicationTwo-stream deep convolutional neural network approach for RGB-D face recognition( 2021-07-21)Two-dimensional face recognition has been researched for the past few decades. With the recent development of Deep Convolutional Neural Network (DCNN) deep learning approaches, two-dimensional face recognition had achieved impressive recognition accuracy rate. However, there are still some challenges such as pose variation, scene illumination, facial emotions, facial occlusions exist in the two-dimensional face recognition. This problem can be solved by adding the depth images as input as it provides valuable information to help model facial boundaries and understand the global facial layout and provide low-frequency patterns. RGB-D images are more robust compared to RGB images. Unfortunately, the lack of sufficient RGB-D face databases to train the DCNN are the main reason for this research to remain undiscovered. So, in this research, new RGB-D face database is constructed using the Intel RealSense D435 Depth Camera which has 1280 x 720-pixel depth. Twin DCNN streams are developed and trained on RGB images at one stream and Depth images at another stream, and finally combined the output through fusion soft-max layers. The proposed DCNN model shows an accuracy of 95% on a newly constructed RGB-D database.
1 -
PublicationApplication of Deep Neural Network for Gas Source Localization in an Indoor Environment( 2023-01-01)Nowadays, the quality of air in the environment has been impacted by the industry. It is important to make sure our ambient air especially in an indoor environment is clean from contaminating particles or harmful gases. Therefore, the air quality inside the indoor environment should be monitored regularly. One of the major problems, when a particular environment has been contaminated by harmful gases, is finding the source of the emission. If the indoor environment has been contaminated by a harmful source it should be instantly localized and eliminated to prevent severe casualties. In this paper, we propose the utilization of synthetic data generated by the Computational Fluid Dynamic (CFD) approach to train the Deep Neural Network (DNN) model called CFD-DNN to perform gas source localization in an indoor environment. The model is capable to locate the contaminated source within a small area of an indoor environment. A total of 361 datasets with different locations of contaminated source release have been obtained using the CFD approach. The obtained dataset was divided into training and testing datasets. The training dataset was used for the model training process while the testing dataset is fed into the model to test model reliability to predict the gas source location. The Euclidian distance equation was used to measure the distance error between the actual and predicted location of the source. The result shows that the model is capable to locate the gas source within a minimum and maximum error of 0.03m to 0.46m respectively.
1 -
PublicationDeep Neural Network for Localizing Gas Source Based on Gas Distribution Map( 2022-01-01)The dynamic characteristic of gas dispersal in uncontrolled environment always leads to inaccurate gas source localization prediction from gas distribution map. Gas distribution map is a representation of the gas distribution over an environment which helps human to observe the concentration of harmful gases at a contaminated area. This paper proposes the utilization of Deep Neural Network (DNN) to predict the gas source location in a gas distribution map. DNN learns from the previous gas distribution map data and patterns to generate a model that is able predict location of gas source. The results indicate that DNN is able to accurately predict the location within the range of 0.8 to 2 m from the actual gas source. This finding shows that DNN has a high potential for utilization in gas source localization application.
4 -
PublicationGas Source Localization via Mobile Robot with Gas Distribution Mapping and Deep Neural Network( 2022-01-01)With the growth of artificial intelligence compute technology, the gas source localization problem would be solved by mobile robots equipped with gas sensing system and artificial intelligence compute units. This work presented a feasibility study of deep learning approach towards gas source localization by mobile robots. A deep neural network strategy was developed and incorporated with the Kernel DM+V gas distribution mapping method. The gas source localization work in this paper was performed on a controlled indoor testbed. From this work, it is shown that by incorporating the developed deep neural network model, it may help improved the gas source location prediction accuracy. A comparison of accuracy between Kernel DM+V and the neural network model is also presented to better visualize the improvement.
3 -
PublicationDeep Neural Network for Localizing Gas Source Based on Gas Distribution Map( 2022-01-01)The dynamic characteristic of gas dispersal in uncontrolled environment always leads to inaccurate gas source localization prediction from gas distribution map. Gas distribution map is a representation of the gas distribution over an environment which helps human to observe the concentration of harmful gases at a contaminated area. This paper proposes the utilization of Deep Neural Network (DNN) to predict the gas source location in a gas distribution map. DNN learns from the previous gas distribution map data and patterns to generate a model that is able predict location of gas source. The results indicate that DNN is able to accurately predict the location within the range of 0.8 to 2 m from the actual gas source. This finding shows that DNN has a high potential for utilization in gas source localization application.
1 -
PublicationGas Source Localization via Mobile Robot with Gas Distribution Mapping and Deep Neural Network( 2022-01-01)With the growth of artificial intelligence compute technology, the gas source localization problem would be solved by mobile robots equipped with gas sensing system and artificial intelligence compute units. This work presented a feasibility study of deep learning approach towards gas source localization by mobile robots. A deep neural network strategy was developed and incorporated with the Kernel DM+V gas distribution mapping method. The gas source localization work in this paper was performed on a controlled indoor testbed. From this work, it is shown that by incorporating the developed deep neural network model, it may help improved the gas source location prediction accuracy. A comparison of accuracy between Kernel DM+V and the neural network model is also presented to better visualize the improvement.
1 -
Publication2D LiDAR Based Reinforcement Learning for Multi-Target Path Planning in Unknown Environment( 2023-01-01)Global path planning techniques have been widely employed in solving path planning problems, however they have been found to be unsuitable for unknown environments. Contrarily, the traditional Q-learning method, which is a common reinforcement learning approach for local path planning, is unable to complete the task for multiple targets. To address these limitations, this paper proposes a modified Q-learning method, called Vector Field Histogram based Q-learning (VFH-QL) utilized the VFH information in state space representation and reward function, based on a 2D LiDAR sensor. We compared the performance of our proposed method with the classical Q-learning method (CQL) through training experiments that were conducted in a simulated environment with a size of 400 square pixels, representing a 20-meter square map. The environment contained static obstacles and a single mobile robot. Two experiments were conducted: experiment A involved path planning for a single target, while experiment B involved path planning for multiple targets. The results of experiment A showed that VFH-QL method had 87.06% less training time and 99.98% better obstacle avoidance compared to CQL. In experiment B, VFH-QL method was found to have an average training time that was 95.69% less than that of the CQL method and 83.99% better path quality. The VFH-QL method was then evaluated using a benchmark dataset. The results indicated that the VFH-QL exhibited superior path quality, with efficiency of 94.89% and improvements of 96.91% and 96.69% over CQL and SARSA in the task of path planning for multiple targets in unknown environments.
1 -
Publication2D LiDAR based reinforcement learning for Multi-Target path planning in unknown environment( 2023)Global path planning techniques have been widely employed in solving path planning problems, however they have been found to be unsuitable for unknown environments. Contrarily, the traditional Q-learning method, which is a common reinforcement learning approach for local path planning, is unable to complete the task for multiple targets. To address these limitations, this paper proposes a modified Q-learning method, called Vector Field Histogram based Q-learning (VFH-QL) utilized the VFH information in state space representation and reward function, based on a 2D LiDAR sensor. We compared the performance of our proposed method with the classical Q-learning method (CQL) through training experiments that were conducted in a simulated environment with a size of 400 square pixels, representing a 20-meter square map. The environment contained static obstacles and a single mobile robot. Two experiments were conducted: experiment A involved path planning for a single target, while experiment B involved path planning for multiple targets. The results of experiment A showed that VFH-QL method had 87.06% less training time and 99.98% better obstacle avoidance compared to CQL. In experiment B, VFH-QL method was found to have an average training time that was 95.69% less than that of the CQL method and 83.99% better path quality. The VFH-QL method was then evaluated using a benchmark dataset. The results indicated that the VFH-QL exhibited superior path quality, with efficiency of 94.89% and improvements of 96.91% and 96.69% over CQL and SARSA in the task of path planning for multiple targets in unknown environments.
11 26 -
PublicationImproved mobile robot based gas distribution mapping through propagated distance transform for structured indoor environment( 2020-05-18)Mobile robot carrying gas sensors have been widely used in mobile olfaction applications. One of the challenging tasks in this research field is Gas Distribution Mapping (GDM). GDM is a representation of how volatile organic compound is spatially dispersed within an environment. This paper addresses the effect of obstacles towards GDM for indoor environment. This work proposes a solution by improvising the Kernel DM + V technique using propagated distance transform (DT) as the weighing function. Since DT computations are CPU heavy, parallel computing, using Compute Unified Device Architecture (CUDA) available in Graphics Processing Unit (GPU), is used to accelerate the DT computation. The proposed solution is compared with the Kernel DM + V algorithm, presenting that the proposed method drastically improves the quality of GDM under various kernel sizes. The study is also further extended towards the effect of obstacles on gas source localization task. The outcome of this work proves that the proposed method shows better accuracy for GDM estimation and gas source localization if obstacle information is considered.
1