TY  - CHAP
AU  - Petrović, Milica
AU  - Jokić, Aleksandar
AU  - Kulesza, Z.
AU  - Miljković, Zoran
PY  - 2021
UR  - https://machinery.mas.bg.ac.rs/handle/123456789/3661
AB  - In the last two decades, the development of state -of-the-art artificial intelligence (AI) models has significantly increased the utilization of commercial and task-specific robots in the service domain. The additional level of intelligence introduced by AI models has enabled service robots to coexist within different human environments and collaborate with end-users. One of the most promising AI techniques, Deep Learning (DL), can provide service robots with a wide range of abilities, such as detecting human pose and emotions, understanding natural languages, as well as scene understanding. Achieved abilities can enable mobile service robots to execute specific tasks in real and stochastic environments. Having that in mind, in this chapter, we provide an in-depth analysis of the tasks that are best-suited for DL within the service robots domain. Moreover, the study of the state-of-the-art DL models for object detection, semantic segmentation, and human pose estimation is carried out. In the end, the authors presented a thorough examination of the training process and analysis of the results for one of the most promising convolutional neural network models (DeepLabv3+) used for semantic segmentation.
PB  - Nova Science Publishers, Inc.
T2  - Service Robots: Advances in Research and Applications
T1  - Deep learning of mobile service robots
EP  - 97
SP  - 77
UR  - https://hdl.handle.net/21.15107/rcub_machinery_3661
ER  - 

@inbook{
author = "Petrović, Milica and Jokić, Aleksandar and Kulesza, Z. and Miljković, Zoran",
year = "2021",
abstract = "In the last two decades, the development of state -of-the-art artificial intelligence (AI) models has significantly increased the utilization of commercial and task-specific robots in the service domain. The additional level of intelligence introduced by AI models has enabled service robots to coexist within different human environments and collaborate with end-users. One of the most promising AI techniques, Deep Learning (DL), can provide service robots with a wide range of abilities, such as detecting human pose and emotions, understanding natural languages, as well as scene understanding. Achieved abilities can enable mobile service robots to execute specific tasks in real and stochastic environments. Having that in mind, in this chapter, we provide an in-depth analysis of the tasks that are best-suited for DL within the service robots domain. Moreover, the study of the state-of-the-art DL models for object detection, semantic segmentation, and human pose estimation is carried out. In the end, the authors presented a thorough examination of the training process and analysis of the results for one of the most promising convolutional neural network models (DeepLabv3+) used for semantic segmentation.",
publisher = "Nova Science Publishers, Inc.",
journal = "Service Robots: Advances in Research and Applications",
booktitle = "Deep learning of mobile service robots",
pages = "97-77",
url = "https://hdl.handle.net/21.15107/rcub_machinery_3661"
}

Petrović, M., Jokić, A., Kulesza, Z.,& Miljković, Z.. (2021). Deep learning of mobile service robots. in Service Robots: Advances in Research and Applications
Nova Science Publishers, Inc.., 77-97.
https://hdl.handle.net/21.15107/rcub_machinery_3661

Petrović M, Jokić A, Kulesza Z, Miljković Z. Deep learning of mobile service robots. in Service Robots: Advances in Research and Applications. 2021;:77-97.
https://hdl.handle.net/21.15107/rcub_machinery_3661 .

Petrović, Milica, Jokić, Aleksandar, Kulesza, Z., Miljković, Zoran, "Deep learning of mobile service robots" in Service Robots: Advances in Research and Applications (2021):77-97,
https://hdl.handle.net/21.15107/rcub_machinery_3661 .

TY  - CONF
AU  - Jokić, Aleksandar
AU  - Petrović, Milica
AU  - Kulesza, Z.
AU  - Miljković, Zoran
PY  - 2021
UR  - https://machinery.mas.bg.ac.rs/handle/123456789/3649
AB  - The recent development of faster and more accurate deep learning models has enabled researchers to utilize the potential of deep learning in robotics. Convolutional neural networks used for the process of semantic segmentation are being applied to improve the traditional robotic tasks by adding an additional level of intelligence, through the execution of context-aware tasks. Having that in mind, visual servoing can now be performed in a completely new manner, by exploiting only semantic and geometric knowledge about the environment. To carry out visual servoing, the mathematical model of the error between the images generated at the current and the desired mobile robot pose (i.e. position and orientation) in the image space needs to be adequately defined. In this paper, we propose the novel mathematical model for the weighted fitness function evaluation, which is utilized for the image registration process within the visual servoing framework. By weighting the classes by their importance in the desired image, the convergence domain of the initial error in the visual servoing process can be greatly extended. The experimental evaluation is carried out on the mobile robot RAICO (Robot with Artificial Intelligence based COgnition), where it is shown that weighted fitness function enables more robust intelligent visual servoing systems with a lower possibility of failure, easier real-world implementation, and feasible object driven navigation.
PB  - Springer Science and Business Media Deutschland GmbH
C3  - Lecture Notes in Networks and Systems
T1  - Visual Deep Learning-Based Mobile Robot Control: A Novel Weighted Fitness Function-Based Image Registration Model
EP  - 752
SP  - 744
VL  - 233
DO  - 10.1007/978-3-030-75275-0_82
ER  - 

@conference{
author = "Jokić, Aleksandar and Petrović, Milica and Kulesza, Z. and Miljković, Zoran",
year = "2021",
abstract = "The recent development of faster and more accurate deep learning models has enabled researchers to utilize the potential of deep learning in robotics. Convolutional neural networks used for the process of semantic segmentation are being applied to improve the traditional robotic tasks by adding an additional level of intelligence, through the execution of context-aware tasks. Having that in mind, visual servoing can now be performed in a completely new manner, by exploiting only semantic and geometric knowledge about the environment. To carry out visual servoing, the mathematical model of the error between the images generated at the current and the desired mobile robot pose (i.e. position and orientation) in the image space needs to be adequately defined. In this paper, we propose the novel mathematical model for the weighted fitness function evaluation, which is utilized for the image registration process within the visual servoing framework. By weighting the classes by their importance in the desired image, the convergence domain of the initial error in the visual servoing process can be greatly extended. The experimental evaluation is carried out on the mobile robot RAICO (Robot with Artificial Intelligence based COgnition), where it is shown that weighted fitness function enables more robust intelligent visual servoing systems with a lower possibility of failure, easier real-world implementation, and feasible object driven navigation.",
publisher = "Springer Science and Business Media Deutschland GmbH",
journal = "Lecture Notes in Networks and Systems",
title = "Visual Deep Learning-Based Mobile Robot Control: A Novel Weighted Fitness Function-Based Image Registration Model",
pages = "752-744",
volume = "233",
doi = "10.1007/978-3-030-75275-0_82"
}

Jokić, A., Petrović, M., Kulesza, Z.,& Miljković, Z.. (2021). Visual Deep Learning-Based Mobile Robot Control: A Novel Weighted Fitness Function-Based Image Registration Model. in Lecture Notes in Networks and Systems
Springer Science and Business Media Deutschland GmbH., 233, 744-752.
https://doi.org/10.1007/978-3-030-75275-0_82

Jokić A, Petrović M, Kulesza Z, Miljković Z. Visual Deep Learning-Based Mobile Robot Control: A Novel Weighted Fitness Function-Based Image Registration Model. in Lecture Notes in Networks and Systems. 2021;233:744-752.
doi:10.1007/978-3-030-75275-0_82 .

Jokić, Aleksandar, Petrović, Milica, Kulesza, Z., Miljković, Zoran, "Visual Deep Learning-Based Mobile Robot Control: A Novel Weighted Fitness Function-Based Image Registration Model" in Lecture Notes in Networks and Systems, 233 (2021):744-752,
https://doi.org/10.1007/978-3-030-75275-0_82 . .