Scientific Publications
Peer reviewed scientific articles
List of peer reviewed publications
As an academic scholar, dissemination of my research findings are an important activity. For this, I focus on top tier peer reviewed scientific articles. Below you will find a list of all my peer reviewed scientific publications including conference proceedings, journal articles and books.
2023
Struckmeier, Oliver; Tiwari, Kshitij; Kyrki, Ville
Autoencoding slow representations for semi-supervised data-efficient regression Journal Article
In: Machine Learning, pp. 1–19, 2023.
Abstract | Links | BibTeX | Tags: Bioinspired
@article{struckmeier2023autoencoding,
title = {Autoencoding slow representations for semi-supervised data-efficient regression},
author = {Oliver Struckmeier and Kshitij Tiwari and Ville Kyrki},
url = {https://link.springer.com/article/10.1007/s10994-022-06299-1},
doi = {10.1007/s10994-022-06299-1},
year = {2023},
date = {2023-01-01},
urldate = {2023-01-01},
journal = {Machine Learning},
pages = {1--19},
publisher = {Springer},
abstract = {The slowness principle is a concept inspired by the visual cortex of the brain. It postulates that the underlying generative factors of a quickly varying sensory signal change on a different, slower time scale. By applying this principle to state-of-the-art unsupervised representation learning methods one can learn a latent embedding to perform supervised downstream regression tasks more data efficient. In this paper, we compare different approaches to unsupervised slow representation learning such as Lp
norm based slowness regularization and the SlowVAE, and propose a new term based on Brownian motion used in our method, the S-VAE. We empirically evaluate these slowness regularization terms with respect to their downstream task performance and data efficiency in state estimation and behavioral cloning tasks. We find that slow representations show great performance improvements in settings where only sparse labeled training data is available. Furthermore, we present a theoretical and empirical comparison of the discussed slowness regularization terms. Finally, we discuss how the Fréchet Inception Distance (FID), commonly used to determine the generative capabilities of GANs, can predict the performance of trained models in supervised downstream tasks.},
keywords = {Bioinspired},
pubstate = {published},
tppubtype = {article}
}
The slowness principle is a concept inspired by the visual cortex of the brain. It postulates that the underlying generative factors of a quickly varying sensory signal change on a different, slower time scale. By applying this principle to state-of-the-art unsupervised representation learning methods one can learn a latent embedding to perform supervised downstream regression tasks more data efficient. In this paper, we compare different approaches to unsupervised slow representation learning such as Lp
norm based slowness regularization and the SlowVAE, and propose a new term based on Brownian motion used in our method, the S-VAE. We empirically evaluate these slowness regularization terms with respect to their downstream task performance and data efficiency in state estimation and behavioral cloning tasks. We find that slow representations show great performance improvements in settings where only sparse labeled training data is available. Furthermore, we present a theoretical and empirical comparison of the discussed slowness regularization terms. Finally, we discuss how the Fréchet Inception Distance (FID), commonly used to determine the generative capabilities of GANs, can predict the performance of trained models in supervised downstream tasks.
norm based slowness regularization and the SlowVAE, and propose a new term based on Brownian motion used in our method, the S-VAE. We empirically evaluate these slowness regularization terms with respect to their downstream task performance and data efficiency in state estimation and behavioral cloning tasks. We find that slow representations show great performance improvements in settings where only sparse labeled training data is available. Furthermore, we present a theoretical and empirical comparison of the discussed slowness regularization terms. Finally, we discuss how the Fréchet Inception Distance (FID), commonly used to determine the generative capabilities of GANs, can predict the performance of trained models in supervised downstream tasks.
2022
Routray, Prasanna Kumar; Kanade, Aditya Sanjiv; Tiwari, Kshitij; Pounds, Pauline; Muniyandi, Manivannan
Towards multidimensional textural perception and classification through whisker Proceedings Article
In: 2022 IEEE International Symposium on Robotic and Sensors Environments (ROSE), pp. 1–7, IEEE 2022.
Links | BibTeX | Tags: Bioinspired, Touch Sensing
@inproceedings{routray2022towards,
title = {Towards multidimensional textural perception and classification through whisker},
author = {Prasanna Kumar Routray and Aditya Sanjiv Kanade and Kshitij Tiwari and Pauline Pounds and Manivannan Muniyandi},
url = {https://ieeexplore.ieee.org/abstract/document/9977409},
year = {2022},
date = {2022-01-01},
urldate = {2022-01-01},
booktitle = {2022 IEEE International Symposium on Robotic and Sensors Environments (ROSE)},
pages = {1--7},
organization = {IEEE},
keywords = {Bioinspired, Touch Sensing},
pubstate = {published},
tppubtype = {inproceedings}
}
Tiwari, Kshitij; Sakcak, Basak; Routray, Prasanna; Manivannan, M; LaValle, Steven M
Visibility-inspired models of touch sensors for navigation Proceedings Article
In: 2022 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pp. 13151–13158, IEEE 2022.
Links | BibTeX | Tags: Bioinspired, Touch Sensing
@inproceedings{tiwari2022visibilityb,
title = {Visibility-inspired models of touch sensors for navigation},
author = {Kshitij Tiwari and Basak Sakcak and Prasanna Routray and M Manivannan and Steven M LaValle},
url = {https://ieeexplore.ieee.org/abstract/document/9981084},
year = {2022},
date = {2022-01-01},
urldate = {2022-01-01},
booktitle = {2022 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)},
pages = {13151--13158},
organization = {IEEE},
keywords = {Bioinspired, Touch Sensing},
pubstate = {published},
tppubtype = {inproceedings}
}
2021
Pearson, Martin J; Dora, Shirin; Struckmeier, Oliver; Knowles, Thomas C; Mitchinson, Ben; Tiwari, Kshitij; Kyrki, Ville; Bohte, Sander; Pennartz, Cyriel MA
Multimodal Representation Learning for Place Recognition Using Deep Hebbian Predictive Coding Journal Article
In: Frontiers in Robotics and AI, vol. 8, 2021.
Abstract | Links | BibTeX | Tags: Bioinspired, Touch Sensing
@article{pearson2021multimodal,
title = {Multimodal Representation Learning for Place Recognition Using Deep Hebbian Predictive Coding},
author = {Martin J Pearson and Shirin Dora and Oliver Struckmeier and Thomas C Knowles and Ben Mitchinson and Kshitij Tiwari and Ville Kyrki and Sander Bohte and Cyriel MA Pennartz},
url = {https://www.frontiersin.org/articles/10.3389/frobt.2021.732023/full},
doi = {10.3389/frobt.2021.732023},
year = {2021},
date = {2021-01-01},
urldate = {2021-01-01},
journal = {Frontiers in Robotics and AI},
volume = {8},
publisher = {Frontiers Media SA},
abstract = {Recognising familiar places is a competence required in many engineering applications that interact with the real world such as robot navigation. Combining information from different sensory sources promotes robustness and accuracy of place recognition. However, mismatch in data registration, dimensionality, and timing between modalities remain challenging problems in multisensory place recognition. Spurious data generated by sensor drop-out in multisensory environments is particularly problematic and often resolved through adhoc and brittle solutions. An effective approach to these problems is demonstrated by animals as they gracefully move through the world. Therefore, we take a neuro-ethological approach by adopting self-supervised representation learning based on a neuroscientific model of visual cortex known as predictive coding. We demonstrate how this parsimonious network algorithm which is trained using a local learning rule can be extended to combine visual and tactile sensory cues from a biomimetic robot as it naturally explores a visually aliased environment. The place recognition performance obtained using joint latent representations generated by the network is significantly better than contemporary representation learning techniques. Further, we see evidence of improved robustness at place recognition in face of unimodal sensor drop-out. The proposed multimodal deep predictive coding algorithm presented is also linearly extensible to accommodate more than two sensory modalities, thereby providing an intriguing example of the value of neuro-biologically plausible representation learning for multimodal navigation.},
keywords = {Bioinspired, Touch Sensing},
pubstate = {published},
tppubtype = {article}
}
Recognising familiar places is a competence required in many engineering applications that interact with the real world such as robot navigation. Combining information from different sensory sources promotes robustness and accuracy of place recognition. However, mismatch in data registration, dimensionality, and timing between modalities remain challenging problems in multisensory place recognition. Spurious data generated by sensor drop-out in multisensory environments is particularly problematic and often resolved through adhoc and brittle solutions. An effective approach to these problems is demonstrated by animals as they gracefully move through the world. Therefore, we take a neuro-ethological approach by adopting self-supervised representation learning based on a neuroscientific model of visual cortex known as predictive coding. We demonstrate how this parsimonious network algorithm which is trained using a local learning rule can be extended to combine visual and tactile sensory cues from a biomimetic robot as it naturally explores a visually aliased environment. The place recognition performance obtained using joint latent representations generated by the network is significantly better than contemporary representation learning techniques. Further, we see evidence of improved robustness at place recognition in face of unimodal sensor drop-out. The proposed multimodal deep predictive coding algorithm presented is also linearly extensible to accommodate more than two sensory modalities, thereby providing an intriguing example of the value of neuro-biologically plausible representation learning for multimodal navigation.
Shukla, Rishabh; Routray, Prasanna Kumar; Tiwari, Kshitij; LaValle, Steven M; Manivannan, M
Monofilament whisker-based mobile robot navigation Proceedings Article
In: 2021 IEEE World Haptics Conference (WHC), pp. 1150–1150, IEEE IEEE, 2021.
Abstract | Links | BibTeX | Tags: Bioinspired
@inproceedings{shukla2021monofilament,
title = {Monofilament whisker-based mobile robot navigation},
author = {Rishabh Shukla and Prasanna Kumar Routray and Kshitij Tiwari and Steven M LaValle and M Manivannan},
url = {https://ieeexplore.ieee.org/document/9517235},
doi = {10.1109/WHC49131.2021.9517235},
year = {2021},
date = {2021-01-01},
urldate = {2021-01-01},
booktitle = {2021 IEEE World Haptics Conference (WHC)},
pages = {1150--1150},
publisher = {IEEE},
organization = {IEEE},
abstract = {Rodents and pinnipeds use their tactile sense as the primary modes for localization and foraging. This abstract presents a rodent whisker-inspired tactile sensor for mobile robot navigation fabricated using the Semmes-Weinstein Monofilament (SWMF). The monofilament is lightweight, readily available, and generates reproducible buckling stresses. The sensor deflection characteristic has three distinct regions of operation that help detect buckling. The sensor’s small size, 360° tactile view, and force sensing capability could make it suitable for narrow space navigation.},
keywords = {Bioinspired},
pubstate = {published},
tppubtype = {inproceedings}
}
Rodents and pinnipeds use their tactile sense as the primary modes for localization and foraging. This abstract presents a rodent whisker-inspired tactile sensor for mobile robot navigation fabricated using the Semmes-Weinstein Monofilament (SWMF). The monofilament is lightweight, readily available, and generates reproducible buckling stresses. The sensor deflection characteristic has three distinct regions of operation that help detect buckling. The sensor’s small size, 360° tactile view, and force sensing capability could make it suitable for narrow space navigation.
Tiwari, Kshitij; Kyrki, Ville; Cheung, Allen; Yamamoto, Naohide
DeFINE: Delayed feedback-based immersive navigation environment for studying goal-directed human navigation Journal Article
In: Behavior Research Methods, vol. 53, no. 6, pp. 2668–2688, 2021.
Abstract | Links | BibTeX | Tags: Bioinspired
@article{tiwari2021define,
title = {DeFINE: Delayed feedback-based immersive navigation environment for studying goal-directed human navigation},
author = {Kshitij Tiwari and Ville Kyrki and Allen Cheung and Naohide Yamamoto},
url = {https://link.springer.com/article/10.3758/s13428-021-01586-6},
doi = {10.3758/s13428-021-01586-6},
year = {2021},
date = {2021-01-01},
urldate = {2021-01-01},
journal = {Behavior Research Methods},
volume = {53},
number = {6},
pages = {2668--2688},
publisher = {Springer},
abstract = {With the advent of consumer-grade products for presenting an immersive virtual environment (VE), there is a growing interest in utilizing VEs for testing human navigation behavior. However, preparing a VE still requires a high level of technical expertise in computer graphics and virtual reality, posing a significant hurdle to embracing the emerging technology. To address this issue, this paper presents Delayed Feedback-based Immersive Navigation Environment (DeFINE), a framework that allows for easy creation and administration of navigation tasks within customizable VEs via intuitive graphical user interfaces and simple settings files. Importantly, DeFINE has a built-in capability to provide performance feedback to participants during an experiment, a feature that is critically missing in other similar frameworks. To show the usability of DeFINE from both experimentalists’ and participants’ perspectives, a demonstration was made in which participants navigated to a hidden goal location with feedback that differentially weighted speed and accuracy of their responses. In addition, the participants evaluated DeFINE in terms of its ease of use, required workload, and proneness to induce cybersickness. The demonstration exemplified typical experimental manipulations DeFINE accommodates and what types of data it can collect for characterizing participants’ task performance. With its out-of-the-box functionality and potential customizability due to open-source licensing, DeFINE makes VEs more accessible to many researchers.
},
keywords = {Bioinspired},
pubstate = {published},
tppubtype = {article}
}
With the advent of consumer-grade products for presenting an immersive virtual environment (VE), there is a growing interest in utilizing VEs for testing human navigation behavior. However, preparing a VE still requires a high level of technical expertise in computer graphics and virtual reality, posing a significant hurdle to embracing the emerging technology. To address this issue, this paper presents Delayed Feedback-based Immersive Navigation Environment (DeFINE), a framework that allows for easy creation and administration of navigation tasks within customizable VEs via intuitive graphical user interfaces and simple settings files. Importantly, DeFINE has a built-in capability to provide performance feedback to participants during an experiment, a feature that is critically missing in other similar frameworks. To show the usability of DeFINE from both experimentalists’ and participants’ perspectives, a demonstration was made in which participants navigated to a hidden goal location with feedback that differentially weighted speed and accuracy of their responses. In addition, the participants evaluated DeFINE in terms of its ease of use, required workload, and proneness to induce cybersickness. The demonstration exemplified typical experimental manipulations DeFINE accommodates and what types of data it can collect for characterizing participants’ task performance. With its out-of-the-box functionality and potential customizability due to open-source licensing, DeFINE makes VEs more accessible to many researchers.
2019
Tiwari, Kshitij; Chong, Nak Young
Multi-Robot Exploration for Environmental Monitoring: The Resource Constrained Perspective Book
Elsevier Inc., 2019, ISBN: 978-0-12-817607-8.
Abstract | Links | BibTeX | Tags: Resource Constraint
@book{tiwari2019multi,
title = {Multi-Robot Exploration for Environmental Monitoring: The Resource Constrained Perspective},
author = {Kshitij Tiwari and Nak Young Chong},
url = {https://www.sciencedirect.com/book/9780128176078/multi-robot-exploration-for-environmental-monitoring;
https://www.amazon.com/gp/product/B0825SX2V7/ref=dbs_a_def_rwt_hsch_vapi_tkin_p1_i0},
doi = {10.1016/C2018-0-01886-1},
isbn = {978-0-12-817607-8},
year = {2019},
date = {2019-11-29},
urldate = {2019-11-29},
publisher = {Elsevier Inc.},
abstract = {Multi-robot Exploration for Environmental Monitoring: The Resource Constrained Perspective provides readers with the necessary robotics and mathematical tools required to realize the correct architecture. The architecture discussed in the book is not confined to environment monitoring, but can also be extended to search-and-rescue, border patrolling, crowd management and related applications. Several law enforcement agencies have already started to deploy UAVs, but instead of using teleoperated UAVs this book proposes methods to fully automate surveillance missions. Similarly, several government agencies like the US-EPA can benefit from this book by automating the process.
Several challenges when deploying such models in real missions are addressed and solved, thus laying stepping stones towards realizing the architecture proposed. This book will be a great resource for graduate students in Computer Science, Computer Engineering, Robotics, Machine Learning and Mechatronics.},
keywords = {Resource Constraint},
pubstate = {published},
tppubtype = {book}
}
Multi-robot Exploration for Environmental Monitoring: The Resource Constrained Perspective provides readers with the necessary robotics and mathematical tools required to realize the correct architecture. The architecture discussed in the book is not confined to environment monitoring, but can also be extended to search-and-rescue, border patrolling, crowd management and related applications. Several law enforcement agencies have already started to deploy UAVs, but instead of using teleoperated UAVs this book proposes methods to fully automate surveillance missions. Similarly, several government agencies like the US-EPA can benefit from this book by automating the process.
Several challenges when deploying such models in real missions are addressed and solved, thus laying stepping stones towards realizing the architecture proposed. This book will be a great resource for graduate students in Computer Science, Computer Engineering, Robotics, Machine Learning and Mechatronics.
Several challenges when deploying such models in real missions are addressed and solved, thus laying stepping stones towards realizing the architecture proposed. This book will be a great resource for graduate students in Computer Science, Computer Engineering, Robotics, Machine Learning and Mechatronics.
Struckmeier, Oliver; Tiwari, Kshitij; Salman, Mohammed; Pearson, Martin J; Kyrki, Ville
Vita-slam: A bio-inspired visuo-tactile slam for navigation while interacting with aliased environments Proceedings Article
In: 2019 IEEE International Conference on Cyborg and Bionic Systems (CBS), pp. 97–103, IEEE 2019.
Abstract | Links | BibTeX | Tags: Bioinspired
@inproceedings{struckmeier2019vita,
title = {Vita-slam: A bio-inspired visuo-tactile slam for navigation while interacting with aliased environments},
author = {Oliver Struckmeier and Kshitij Tiwari and Mohammed Salman and Martin J Pearson and Ville Kyrki},
url = {https://ieeexplore.ieee.org/abstract/document/9114526},
doi = {10.1109/CBS46900.2019.9114526},
year = {2019},
date = {2019-01-01},
urldate = {2019-01-01},
booktitle = {2019 IEEE International Conference on Cyborg and Bionic Systems (CBS)},
pages = {97--103},
organization = {IEEE},
abstract = {RatSLAM is a rat hippocampus-inspired visual Simultaneous Localization and Mapping (SLAM) framework capable of generating semi-metric topological representations of indoor and outdoor environments. Whisker-RatSLAM is a 6D extension of the RatSLAM and primarily focuses on object recognition by generating point clouds of objects based on tactile information from an array of biomimetic whiskers. This paper introduces a novel extension to both former works that is referred to as ViTa-SLAM that harnesses both vision and tactile information for performing SLAM. This not only allows the robot to perform natural interactions with the environment whilst navigating, as is normally seen in nature, but also provides a mechanism to fuse non-unique tactile and unique visual data. Compared to the former works, our approach can handle ambiguous scenes in which one sensor alone is not capable of identifying false-positive loop-closures.},
keywords = {Bioinspired},
pubstate = {published},
tppubtype = {inproceedings}
}
RatSLAM is a rat hippocampus-inspired visual Simultaneous Localization and Mapping (SLAM) framework capable of generating semi-metric topological representations of indoor and outdoor environments. Whisker-RatSLAM is a 6D extension of the RatSLAM and primarily focuses on object recognition by generating point clouds of objects based on tactile information from an array of biomimetic whiskers. This paper introduces a novel extension to both former works that is referred to as ViTa-SLAM that harnesses both vision and tactile information for performing SLAM. This not only allows the robot to perform natural interactions with the environment whilst navigating, as is normally seen in nature, but also provides a mechanism to fuse non-unique tactile and unique visual data. Compared to the former works, our approach can handle ambiguous scenes in which one sensor alone is not capable of identifying false-positive loop-closures.
Tiwari, Kshitij; Xiao, Xuesu; Malik, Ashish; Chong, Nak Young
A unified framework for operational range estimation of mobile robots operating on a single discharge to avoid complete immobilization Journal Article
In: Mechatronics, vol. 57, pp. 173–187, 2019.
Abstract | Links | BibTeX | Tags: Resource Constraint
@article{tiwari2019unified,
title = {A unified framework for operational range estimation of mobile robots operating on a single discharge to avoid complete immobilization},
author = {Kshitij Tiwari and Xuesu Xiao and Ashish Malik and Nak Young Chong},
url = {https://www.sciencedirect.com/science/article/pii/S0957415818301946?casa_token=3NPC8KewBfsAAAAA:F1aDckAolYCwgofhCdnFtZg6L6aRKboA-T1z_X5ZmkV_O8AFVy5-NW3fsV0bk8BxtBMlcHxp2g},
doi = {10.1016/j.mechatronics.2018.12.006},
year = {2019},
date = {2019-01-01},
urldate = {2019-01-01},
journal = {Mechatronics},
volume = {57},
pages = {173--187},
publisher = {Elsevier},
abstract = {Mobile robots are being increasingly deployed in fields where human intervention is deemed risky. However, in doing so, one of the prime concern is to prevent complete battery depletion which may in turn lead to immobilization of the robot during the mission. Thus, we need to carefully manage the energy available to explore as much of the unknown environment as feasible whilst guaranteeing a safe return journey to home base. For this, we need to identify the key components that draw energy and quantify their individual energy requirements. However, this problem is difficult due to the fact that most of the robots have different motion models, and the energy consumption usually also varies from mission to mission. It is desirable to have a generic framework that takes into account different locomotion models and possible mission profiles. This paper presents a methodology to unify the energy consumption models for various robotic platforms thereby allowing us to estimate operational range in both offline and online fashions. The existing models consider a given mission profile and try to estimate its energy requirements whilst our model considers the energy as a given resource constraint and tries to optimize the mission to be accomplished within these constraints. The proposed unified energy consumption framework is verified by field experiments for micro UGV and multi-rotor UAV test-beds operating under myriad of environmental conditions. The online model estimates operational range with an average accuracy (measured with respect to true range across multiple field trials) of 93.87% while the offline model attains 82.97%.},
keywords = {Resource Constraint},
pubstate = {published},
tppubtype = {article}
}
Mobile robots are being increasingly deployed in fields where human intervention is deemed risky. However, in doing so, one of the prime concern is to prevent complete battery depletion which may in turn lead to immobilization of the robot during the mission. Thus, we need to carefully manage the energy available to explore as much of the unknown environment as feasible whilst guaranteeing a safe return journey to home base. For this, we need to identify the key components that draw energy and quantify their individual energy requirements. However, this problem is difficult due to the fact that most of the robots have different motion models, and the energy consumption usually also varies from mission to mission. It is desirable to have a generic framework that takes into account different locomotion models and possible mission profiles. This paper presents a methodology to unify the energy consumption models for various robotic platforms thereby allowing us to estimate operational range in both offline and online fashions. The existing models consider a given mission profile and try to estimate its energy requirements whilst our model considers the energy as a given resource constraint and tries to optimize the mission to be accomplished within these constraints. The proposed unified energy consumption framework is verified by field experiments for micro UGV and multi-rotor UAV test-beds operating under myriad of environmental conditions. The online model estimates operational range with an average accuracy (measured with respect to true range across multiple field trials) of 93.87% while the offline model attains 82.97%.
2018
Tiwari, Kshitij; Jeong, Sungmoon; Chong, Nak Young
In: IEEE Transactions on Robotics, vol. 34, no. 3, pp. 820–828, 2018.
Abstract | Links | BibTeX | Tags: Resource Constraint
@article{tiwari2018point,
title = {Point-wise fusion of distributed gaussian process experts (fudge) using a fully decentralized robot team operating in communication-devoid environment},
author = {Kshitij Tiwari and Sungmoon Jeong and Nak Young Chong},
url = {https://ieeexplore.ieee.org/document/8304807},
doi = {10.1109/TRO.2018.2794535},
year = {2018},
date = {2018-01-01},
urldate = {2018-01-01},
journal = {IEEE Transactions on Robotics},
volume = {34},
number = {3},
pages = {820--828},
publisher = {IEEE},
abstract = {In this paper, we focus on large-scale environment monitoring by utilizing a fully decentralized team of mobile robots. The robots utilize the resource constrained-decentralized active sensing scheme to select the most informative (uncertain) locations to observe while conserving allocated resources (battery, travel distance, etc.). We utilize a distributed Gaussian process (GP) framework to split the computational load over our fleet of robots. Since each robot is individually generating a model of the environment, there may be conflicting predictions for test locations. Thus, in this paper, we propose an algorithm for aggregating individual prediction models into a single globally consistent model that can be used to infer the overall spatial dynamics of the environment. To make a prediction at a previously unobserved location, we propose a novel gating network for a mixture-of-experts model wherein the weight of an expert is determined by the responsibility of the expert over the unvisited location. The benefit of posing our problem as a centralized fusion with a distributed GP computation approach is that the robots never communicate with each other, individually optimize their own GP models based on their respective observations, and off-load all their learnt models on the base station only at the end of their respective mission times. We demonstrate the effectiveness of our approach using publicly available datasets.},
keywords = {Resource Constraint},
pubstate = {published},
tppubtype = {article}
}
In this paper, we focus on large-scale environment monitoring by utilizing a fully decentralized team of mobile robots. The robots utilize the resource constrained-decentralized active sensing scheme to select the most informative (uncertain) locations to observe while conserving allocated resources (battery, travel distance, etc.). We utilize a distributed Gaussian process (GP) framework to split the computational load over our fleet of robots. Since each robot is individually generating a model of the environment, there may be conflicting predictions for test locations. Thus, in this paper, we propose an algorithm for aggregating individual prediction models into a single globally consistent model that can be used to infer the overall spatial dynamics of the environment. To make a prediction at a previously unobserved location, we propose a novel gating network for a mixture-of-experts model wherein the weight of an expert is determined by the responsibility of the expert over the unvisited location. The benefit of posing our problem as a centralized fusion with a distributed GP computation approach is that the robots never communicate with each other, individually optimize their own GP models based on their respective observations, and off-load all their learnt models on the base station only at the end of their respective mission times. We demonstrate the effectiveness of our approach using publicly available datasets.
Tiwari, Kshitij
Multi-robot resource constrained decentralized active sensing for online environment monitoring PhD Thesis
PhD thesis, School of Information Sci., JAIST, Japan, 2018., 2018.
Abstract | Links | BibTeX | Tags: Resource Constraint
@phdthesis{tiwari2018multi,
title = {Multi-robot resource constrained decentralized active sensing for online environment monitoring},
author = {Kshitij Tiwari},
url = {https://dspace.jaist.ac.jp/dspace/bitstream/10119/15336/6/paper.pdf},
year = {2018},
date = {2018-01-01},
urldate = {2018-01-01},
school = {PhD thesis, School of Information Sci., JAIST, Japan, 2018.},
abstract = {This thesis addresses the problem of trajectory planning over discrete domains
for monitoring an environmental phenomenon that is varying spatially. The most
relevant application corresponds to environmental monitoring using an autonomous
mobile robot for air, water or land pollution monitoring. Since the dynamics of the
phenomenon are not known a priori, the planning algorithm needs to satisfy two
objectives simultaneously: 1) Learn and predict spatial patterns and, 2) adhere to
resource constraints while gathering observations. Subsequently, the thesis brings the
following contributions:
Firstly, it formulates a resource constrained information-theoretic path planning
scheme called Resource Constrained Decentralized Active Sensing (RC-DAS) that can
effectively trade-off model performance to resource utilization. Since, these objectives
are inherently conflicting, optimizing over both these objectives is rather challenging.
However, weighted combination of these objectives into a single objective function is
proposed such that the total path length is bounded by the maximum operational
range. This path planner is then coupled with a distributed Gaussian Process (DGP)
framework to allow the robots to simultaneously infer and predict the dynamics of
the environment of interest.
Secondly, optimal weight selection method is proposed wherein the weights of the
RC-DAS cost function are dynamically updated as a function of residual resources.
This extended scheme is referred to as RC-DAS† which additionally ensures that
the robots return to base station at the end of their respective mission times. This
prevents the robots from getting stranded amidst the field and is a first step towards
making the architecture fail-proof.
Thirdly, an operational range estimation framework is proposed to interpret the
bounds on maximum path length attainable by the robots. This should be used as
the limiting condition for terminating the exploration to ensure a safe path to the
base station. This framework is then generalized to encompass various classes of
robots and is made robust to operate with high accuracy even when subject to natural
environmental disturbances like strong wind gusts or uneven terrains.
Fourthly, the RC-DAS framework is scaled to multiple robots operating in a fully
decentralized fashion in communication devoid environments. Owing to such a setting,
multiple inferred models of the environment can be obtained. However, neither all
models can be fully trusted nor forthrightly rejected. To solve this dilemma and to
obtain one globally consistent model, a pointwise fusion of distributed GP models is
introduced and referred to as FuDGE},
keywords = {Resource Constraint},
pubstate = {published},
tppubtype = {phdthesis}
}
This thesis addresses the problem of trajectory planning over discrete domains
for monitoring an environmental phenomenon that is varying spatially. The most
relevant application corresponds to environmental monitoring using an autonomous
mobile robot for air, water or land pollution monitoring. Since the dynamics of the
phenomenon are not known a priori, the planning algorithm needs to satisfy two
objectives simultaneously: 1) Learn and predict spatial patterns and, 2) adhere to
resource constraints while gathering observations. Subsequently, the thesis brings the
following contributions:
Firstly, it formulates a resource constrained information-theoretic path planning
scheme called Resource Constrained Decentralized Active Sensing (RC-DAS) that can
effectively trade-off model performance to resource utilization. Since, these objectives
are inherently conflicting, optimizing over both these objectives is rather challenging.
However, weighted combination of these objectives into a single objective function is
proposed such that the total path length is bounded by the maximum operational
range. This path planner is then coupled with a distributed Gaussian Process (DGP)
framework to allow the robots to simultaneously infer and predict the dynamics of
the environment of interest.
Secondly, optimal weight selection method is proposed wherein the weights of the
RC-DAS cost function are dynamically updated as a function of residual resources.
This extended scheme is referred to as RC-DAS† which additionally ensures that
the robots return to base station at the end of their respective mission times. This
prevents the robots from getting stranded amidst the field and is a first step towards
making the architecture fail-proof.
Thirdly, an operational range estimation framework is proposed to interpret the
bounds on maximum path length attainable by the robots. This should be used as
the limiting condition for terminating the exploration to ensure a safe path to the
base station. This framework is then generalized to encompass various classes of
robots and is made robust to operate with high accuracy even when subject to natural
environmental disturbances like strong wind gusts or uneven terrains.
Fourthly, the RC-DAS framework is scaled to multiple robots operating in a fully
decentralized fashion in communication devoid environments. Owing to such a setting,
multiple inferred models of the environment can be obtained. However, neither all
models can be fully trusted nor forthrightly rejected. To solve this dilemma and to
obtain one globally consistent model, a pointwise fusion of distributed GP models is
introduced and referred to as FuDGE
for monitoring an environmental phenomenon that is varying spatially. The most
relevant application corresponds to environmental monitoring using an autonomous
mobile robot for air, water or land pollution monitoring. Since the dynamics of the
phenomenon are not known a priori, the planning algorithm needs to satisfy two
objectives simultaneously: 1) Learn and predict spatial patterns and, 2) adhere to
resource constraints while gathering observations. Subsequently, the thesis brings the
following contributions:
Firstly, it formulates a resource constrained information-theoretic path planning
scheme called Resource Constrained Decentralized Active Sensing (RC-DAS) that can
effectively trade-off model performance to resource utilization. Since, these objectives
are inherently conflicting, optimizing over both these objectives is rather challenging.
However, weighted combination of these objectives into a single objective function is
proposed such that the total path length is bounded by the maximum operational
range. This path planner is then coupled with a distributed Gaussian Process (DGP)
framework to allow the robots to simultaneously infer and predict the dynamics of
the environment of interest.
Secondly, optimal weight selection method is proposed wherein the weights of the
RC-DAS cost function are dynamically updated as a function of residual resources.
This extended scheme is referred to as RC-DAS† which additionally ensures that
the robots return to base station at the end of their respective mission times. This
prevents the robots from getting stranded amidst the field and is a first step towards
making the architecture fail-proof.
Thirdly, an operational range estimation framework is proposed to interpret the
bounds on maximum path length attainable by the robots. This should be used as
the limiting condition for terminating the exploration to ensure a safe path to the
base station. This framework is then generalized to encompass various classes of
robots and is made robust to operate with high accuracy even when subject to natural
environmental disturbances like strong wind gusts or uneven terrains.
Fourthly, the RC-DAS framework is scaled to multiple robots operating in a fully
decentralized fashion in communication devoid environments. Owing to such a setting,
multiple inferred models of the environment can be obtained. However, neither all
models can be fully trusted nor forthrightly rejected. To solve this dilemma and to
obtain one globally consistent model, a pointwise fusion of distributed GP models is
introduced and referred to as FuDGE
Tiwari, Kshitij; Xiao, Xuesu; Chong, Nak Young
Estimating achievable range of ground robots operating on single battery discharge for operational efficacy amelioration Proceedings Article
In: 2018 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pp. 3991–3998, IEEE 2018.
Abstract | Links | BibTeX | Tags: Resource Constraint
@inproceedings{tiwari2018estimating,
title = {Estimating achievable range of ground robots operating on single battery discharge for operational efficacy amelioration},
author = {Kshitij Tiwari and Xuesu Xiao and Nak Young Chong},
url = {https://ieeexplore.ieee.org/abstract/document/8593845},
doi = {10.1109/IROS.2018.8593845},
year = {2018},
date = {2018-01-01},
urldate = {2018-01-01},
booktitle = {2018 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)},
pages = {3991--3998},
organization = {IEEE},
abstract = {Mobile robots are increasingly being used to assist with active pursuit and law enforcement. One major limitation for such missions is the resource (battery) allocated to the robot. Factors like nature and agility of evader, terrain over which pursuit is being carried out, plausible traversal velocity and the amount of necessary data to be collected all influence how long the robot can last in the field and how far it can travel. In this paper, we develop an analytical model that analyzes the energy utilization for a variety of components mounted on a robot to estimate the maximum operational range achievable by the robot operating on a single battery discharge. We categorize the major consumers of energy as: 1.) ancillary robotic functions such as computation, communication, sensing etc., and 2.) maneuvering which involves propulsion, steering etc. Both these consumers draw power from the common power source but the achievable range is largely affected by the proportion of power available for maneuvering. For this case study, we performed experiments with real robots on planar and graded surfaces and evaluated the estimation error for each case.},
keywords = {Resource Constraint},
pubstate = {published},
tppubtype = {inproceedings}
}
Mobile robots are increasingly being used to assist with active pursuit and law enforcement. One major limitation for such missions is the resource (battery) allocated to the robot. Factors like nature and agility of evader, terrain over which pursuit is being carried out, plausible traversal velocity and the amount of necessary data to be collected all influence how long the robot can last in the field and how far it can travel. In this paper, we develop an analytical model that analyzes the energy utilization for a variety of components mounted on a robot to estimate the maximum operational range achievable by the robot operating on a single battery discharge. We categorize the major consumers of energy as: 1.) ancillary robotic functions such as computation, communication, sensing etc., and 2.) maneuvering which involves propulsion, steering etc. Both these consumers draw power from the common power source but the achievable range is largely affected by the proportion of power available for maneuvering. For this case study, we performed experiments with real robots on planar and graded surfaces and evaluated the estimation error for each case.
2017
Tiwari, Kshitij; Jeong, Sungmoon; Chong, Nak Young
Map-reduce Gaussian process (MR-GP) for multi-UAV based environment monitoring with limited battery Proceedings Article
In: 2017 56th Annual Conference of the Society of Instrument and Control Engineers of Japan (SICE), pp. 760–763, IEEE 2017.
Abstract | Links | BibTeX | Tags: Resource Constraint
@inproceedings{tiwari2017map,
title = {Map-reduce Gaussian process (MR-GP) for multi-UAV based environment monitoring with limited battery},
author = {Kshitij Tiwari and Sungmoon Jeong and Nak Young Chong},
url = {https://ieeexplore.ieee.org/abstract/document/8105445},
doi = {10.23919/SICE.2017.8105445},
year = {2017},
date = {2017-01-01},
urldate = {2017-01-01},
booktitle = {2017 56th Annual Conference of the Society of Instrument and Control Engineers of Japan (SICE)},
pages = {760--763},
organization = {IEEE},
abstract = {Environment monitoring is a challenging task owing to its ever changing dynamics. Furthermore, deploying a team of resource constrained robots to persistently monitor the environment encompasses intelligently selecting the training samples which are spread across a significantly large area to conservatively spend the resources allocated. In order to accomplish this using a team of fully autonomous self-reliant robots, we pose this problem as a map-reduce architecture: Map phase involves each individual member gathering its training samples and generating the best possible model of the environment followed by the Reduce phase where we merge all these models into a single globally consistent model to infer the environment dynamics. Our preliminary contributions to both these phases have shown significant ease to parallelize the process of gathering training samples whilst reducing the over-all model uncertainty. We demonstrated these results in a communication devoid simulated environment using publicly available datasets.},
keywords = {Resource Constraint},
pubstate = {published},
tppubtype = {inproceedings}
}
Environment monitoring is a challenging task owing to its ever changing dynamics. Furthermore, deploying a team of resource constrained robots to persistently monitor the environment encompasses intelligently selecting the training samples which are spread across a significantly large area to conservatively spend the resources allocated. In order to accomplish this using a team of fully autonomous self-reliant robots, we pose this problem as a map-reduce architecture: Map phase involves each individual member gathering its training samples and generating the best possible model of the environment followed by the Reduce phase where we merge all these models into a single globally consistent model to infer the environment dynamics. Our preliminary contributions to both these phases have shown significant ease to parallelize the process of gathering training samples whilst reducing the over-all model uncertainty. We demonstrated these results in a communication devoid simulated environment using publicly available datasets.
Tiwari, Kshitij; Jeong, Sungmoon; Chong, Nak Young
Multi-uav resource constrained online monitoring of large-scale spatio-temporal environment with homing guarantee Proceedings Article
In: IECON 2017-43rd Annual Conference of the IEEE Industrial Electronics Society, pp. 5893–5900, IEEE IEEE, 2017.
Abstract | Links | BibTeX | Tags: Resource Constraint
@inproceedings{tiwari2017multi,
title = {Multi-uav resource constrained online monitoring of large-scale spatio-temporal environment with homing guarantee},
author = {Kshitij Tiwari and Sungmoon Jeong and Nak Young Chong},
url = {https://ieeexplore.ieee.org/document/8217022},
doi = {10.1109/IECON.2017.8217022},
year = {2017},
date = {2017-01-01},
urldate = {2017-01-01},
booktitle = {IECON 2017-43rd Annual Conference of the IEEE Industrial Electronics Society},
pages = {5893--5900},
publisher = {IEEE},
organization = {IEEE},
abstract = {We propose a homing constrained bi-objective optimization variant of budget-limited informative path planning for monitoring a spatio-temporal environment. The objective function consists of weighted combination of two components: model performance which must be maximized and travel distance which must be bounded by the maximum operational range. Besides this, we have additional constraints that guarantee that the robots will return to home (base station) upon completion of their respective missions. Optimizing over this objective function is essentially NP-hard owing to the conflicting constituents. Moreover, the appropriate choice of weights and additional homing guarantees further adds to complications. We employ Gaussian Process (GP) model [1] which is highly data driven i.e., the larger the amount of training data, the better the model performance. However, owing to limited resources, a robot can only collect a limited amount of training samples. Thus, with the introduction of our bi-objective cost function, it becomes possible to plan budget-limited (e.g., battery, flight time, travel distance etc.) informative tours using autonomous mobile robots to effectively select only the most informative (uncertain) locations from the environment. In this work, we develop an algorithm to autonomously choose the appropriate weights for the components based on available resources while ensuring homing and maintaining model quality. We perform simulations to verify the effectiveness of our proposed objective function on the publicly available Ozone Concentration dataset gathered from USA.},
keywords = {Resource Constraint},
pubstate = {published},
tppubtype = {inproceedings}
}
We propose a homing constrained bi-objective optimization variant of budget-limited informative path planning for monitoring a spatio-temporal environment. The objective function consists of weighted combination of two components: model performance which must be maximized and travel distance which must be bounded by the maximum operational range. Besides this, we have additional constraints that guarantee that the robots will return to home (base station) upon completion of their respective missions. Optimizing over this objective function is essentially NP-hard owing to the conflicting constituents. Moreover, the appropriate choice of weights and additional homing guarantees further adds to complications. We employ Gaussian Process (GP) model [1] which is highly data driven i.e., the larger the amount of training data, the better the model performance. However, owing to limited resources, a robot can only collect a limited amount of training samples. Thus, with the introduction of our bi-objective cost function, it becomes possible to plan budget-limited (e.g., battery, flight time, travel distance etc.) informative tours using autonomous mobile robots to effectively select only the most informative (uncertain) locations from the environment. In this work, we develop an algorithm to autonomously choose the appropriate weights for the components based on available resources while ensuring homing and maintaining model quality. We perform simulations to verify the effectiveness of our proposed objective function on the publicly available Ozone Concentration dataset gathered from USA.
2016
Tiwari, Kshitij; Honorè, Valentin; Jeong, Sungmoon; Chong, Nak Young; Deisenroth, Marc Peter
Resource-constrained decentralized active sensing for multi-robot systems using distributed Gaussian processes Proceedings Article
In: 2016 16th International Conference on Control, Automation and Systems (ICCAS), pp. 13–18, IEEE, 2016.
Abstract | Links | BibTeX | Tags: Resource Constraint
@inproceedings{tiwari2016resource,
title = {Resource-constrained decentralized active sensing for multi-robot systems using distributed Gaussian processes},
author = {Kshitij Tiwari and Valentin Honorè and Sungmoon Jeong and Nak Young Chong and Marc Peter Deisenroth},
url = {https://ieeexplore.ieee.org/document/7832293},
doi = {10.1109/ICCAS.2016.7832293},
year = {2016},
date = {2016-10-27},
urldate = {2017-01-26},
booktitle = {2016 16th International Conference on Control, Automation and Systems (ICCAS)},
pages = {13--18},
publisher = {IEEE},
abstract = {We consider the problem of area coverage for robot teams operating under resource constraints, while modeling spatio-temporal environmental phenomena. The aim of the mobile robot team is to avoid exhaustive search and only visit the most important locations that can improve the prediction accuracy of a spatio-temporal model. We use a Gaussian Process (GP) to model spatially varying and temporally evolving dynamics of the target phenomenon. Each robot of the team is allocated a dedicated search area wherein the robot autonomously optimizes its prediction accuracy. We present this as a Decentralized Computation and Centralized Data Fusion approach wherein the trajectory sampled by the robot is generated using our proposed Resource-Constrained Decentralized Active Sensing (RC-DAS). Since each robot possesses its own independent prediction model, at the end of robot's mission time, we fuse all the prediction models from all robots to have a global model of the spatio-temporal phenomenon. Previously, all robots and GPs needed to be synchronized, such that the GPs can be jointly trained. However, doing so defeats the purpose of a fully decentralized mobile robot team. Thus, we allow the robots to independently gather new measurements and update their model parameters irrespective of other members of the team. To evaluate the performance of our model, we compare the trajectory traced by the robot using active and passive (e.g., nearest neighbor selection) sensing. We compare the performance and cost incurred by a resource constrained optimization with the unconstrained entropy maximization version.},
keywords = {Resource Constraint},
pubstate = {published},
tppubtype = {inproceedings}
}
We consider the problem of area coverage for robot teams operating under resource constraints, while modeling spatio-temporal environmental phenomena. The aim of the mobile robot team is to avoid exhaustive search and only visit the most important locations that can improve the prediction accuracy of a spatio-temporal model. We use a Gaussian Process (GP) to model spatially varying and temporally evolving dynamics of the target phenomenon. Each robot of the team is allocated a dedicated search area wherein the robot autonomously optimizes its prediction accuracy. We present this as a Decentralized Computation and Centralized Data Fusion approach wherein the trajectory sampled by the robot is generated using our proposed Resource-Constrained Decentralized Active Sensing (RC-DAS). Since each robot possesses its own independent prediction model, at the end of robot's mission time, we fuse all the prediction models from all robots to have a global model of the spatio-temporal phenomenon. Previously, all robots and GPs needed to be synchronized, such that the GPs can be jointly trained. However, doing so defeats the purpose of a fully decentralized mobile robot team. Thus, we allow the robots to independently gather new measurements and update their model parameters irrespective of other members of the team. To evaluate the performance of our model, we compare the trajectory traced by the robot using active and passive (e.g., nearest neighbor selection) sensing. We compare the performance and cost incurred by a resource constrained optimization with the unconstrained entropy maximization version.
2014
Tiwari, Kshitij
Co-operative manipulation with robots Masters Thesis
School of Informatics, 2014.
Abstract | Links | BibTeX | Tags: Cobots
@mastersthesis{Tiwari2014cooperative,
title = {Co-operative manipulation with robots},
author = {Kshitij Tiwari},
url = {https://rad.inf.ed.ac.uk/rad-people.html},
year = {2014},
date = {2014-01-01},
urldate = {2014-01-01},
institution = {Institute of Perception, Action and Behaviour (IPAB), University of Edinburgh},
school = {School of Informatics},
abstract = {This MSc thesis aims at presenting the work done in the field of human robot collaborations. In this project, the author(s) aimed to develop a symbiotic setup in which a humanoid robot named Baxter was used in order to help a human disassemble a CPU Chassis. In the given time frame, the task of pulling out a PCB from the chassis was realised. In this project, we utilised two types of grippers to be able to retrieve the PCB. We used a vacuum cup gripper and a electric parallel gripper for this task. By means of this project, we coupled visual servoing tasks with co-operative manipulation in order to give the robots the ability to work in unison with a human worker. While this makes the life of human worker easier, it also means they can improve the efficacy at work and especially in a task like disassembly this would eventually lead to cost-effective disassembly procedures. This project is stepping stone towards the development of a work station whereby the human works with a robot as opposed of current scenarios whereby a robot works completely autonomously without any interaction with humans. This would help us overcome the current limitation of AI whereby the algorithms are not robust enough to allow the robot to perform all tasks by itself.},
keywords = {Cobots},
pubstate = {published},
tppubtype = {mastersthesis}
}
This MSc thesis aims at presenting the work done in the field of human robot collaborations. In this project, the author(s) aimed to develop a symbiotic setup in which a humanoid robot named Baxter was used in order to help a human disassemble a CPU Chassis. In the given time frame, the task of pulling out a PCB from the chassis was realised. In this project, we utilised two types of grippers to be able to retrieve the PCB. We used a vacuum cup gripper and a electric parallel gripper for this task. By means of this project, we coupled visual servoing tasks with co-operative manipulation in order to give the robots the ability to work in unison with a human worker. While this makes the life of human worker easier, it also means they can improve the efficacy at work and especially in a task like disassembly this would eventually lead to cost-effective disassembly procedures. This project is stepping stone towards the development of a work station whereby the human works with a robot as opposed of current scenarios whereby a robot works completely autonomously without any interaction with humans. This would help us overcome the current limitation of AI whereby the algorithms are not robust enough to allow the robot to perform all tasks by itself.