Matching Items (7)
Filtering by
- All Subjects: artificial intelligence
- Genre: Masters Thesis
- Member of: ASU Electronic Theses and Dissertations
Description
Visual navigation is a useful and important task for a variety of applications. As the prevalence of robots increase, there is an increasing need for energy-efficient navigation methods as well. Many aspects of efficient visual navigation algorithms have been implemented in the literature, but there is a lack of work on evaluation of the efficiency of the image sensors. In this thesis, two methods are evaluated: adaptive image sensor quantization for traditional camera pipelines as well as new event-based sensors for low-power computer vision.The first contribution in this thesis is an evaluation of performing varying levels of sensor linear and logarithmic quantization with the task of visual simultaneous localization and mapping (SLAM). This unconventional method can provide efficiency benefits with a trade off between accuracy of the task and energy-efficiency. A new sensor quantization method, gradient-based quantization, is introduced to improve the accuracy of the task. This method only lowers the bit level of parts of the image that are less likely to be important in the SLAM algorithm since lower bit levels signify better energy-efficiency, but worse task accuracy. The third contribution is an evaluation of the efficiency and accuracy of event-based camera intensity representations for the task of optical flow. The results of performing a learning based optical flow are provided for each of five different reconstruction methods along with ablation studies. Lastly, the challenges of an event feature-based SLAM system are presented with results demonstrating the necessity for high quality and high resolution event data. The work in this thesis provides studies useful for examining tradeoffs for an efficient visual navigation system with traditional and event vision sensors. The results of this thesis also provide multiple directions for future work.
ContributorsChristie, Olivia Catherine (Author) / Jayasuriya, Suren (Thesis advisor) / Chakrabarti, Chaitali (Committee member) / Yang, Yezhou (Committee member) / Arizona State University (Publisher)
Created2022
Description
Deep neural networks (DNN) have shown tremendous success in various cognitive tasks, such as image classification, speech recognition, etc. However, their usage on resource-constrained edge devices has been limited due to high computation and large memory requirement.
To overcome these challenges, recent works have extensively investigated model compression techniques such as element-wise sparsity, structured sparsity and quantization. While most of these works have applied these compression techniques in isolation, there have been very few studies on application of quantization and structured sparsity together on a DNN model.
This thesis co-optimizes structured sparsity and quantization constraints on DNN models during training. Specifically, it obtains optimal setting of 2-bit weight and 2-bit activation coupled with 4X structured compression by performing combined exploration of quantization and structured compression settings. The optimal DNN model achieves 50X weight memory reduction compared to floating-point uncompressed DNN. This memory saving is significant since applying only structured sparsity constraints achieves 2X memory savings and only quantization constraints achieves 16X memory savings. The algorithm has been validated on both high and low capacity DNNs and on wide-sparse and deep-sparse DNN models. Experiments demonstrated that deep-sparse DNN outperforms shallow-dense DNN with varying level of memory savings depending on DNN precision and sparsity levels. This work further proposed a Pareto-optimal approach to systematically extract optimal DNN models from a huge set of sparse and dense DNN models. The resulting 11 optimal designs were further evaluated by considering overall DNN memory which includes activation memory and weight memory. It was found that there is only a small change in the memory footprint of the optimal designs corresponding to the low sparsity DNNs. However, activation memory cannot be ignored for high sparsity DNNs.
To overcome these challenges, recent works have extensively investigated model compression techniques such as element-wise sparsity, structured sparsity and quantization. While most of these works have applied these compression techniques in isolation, there have been very few studies on application of quantization and structured sparsity together on a DNN model.
This thesis co-optimizes structured sparsity and quantization constraints on DNN models during training. Specifically, it obtains optimal setting of 2-bit weight and 2-bit activation coupled with 4X structured compression by performing combined exploration of quantization and structured compression settings. The optimal DNN model achieves 50X weight memory reduction compared to floating-point uncompressed DNN. This memory saving is significant since applying only structured sparsity constraints achieves 2X memory savings and only quantization constraints achieves 16X memory savings. The algorithm has been validated on both high and low capacity DNNs and on wide-sparse and deep-sparse DNN models. Experiments demonstrated that deep-sparse DNN outperforms shallow-dense DNN with varying level of memory savings depending on DNN precision and sparsity levels. This work further proposed a Pareto-optimal approach to systematically extract optimal DNN models from a huge set of sparse and dense DNN models. The resulting 11 optimal designs were further evaluated by considering overall DNN memory which includes activation memory and weight memory. It was found that there is only a small change in the memory footprint of the optimal designs corresponding to the low sparsity DNNs. However, activation memory cannot be ignored for high sparsity DNNs.
ContributorsSrivastava, Gaurav (Author) / Seo, Jae-Sun (Thesis advisor) / Chakrabarti, Chaitali (Committee member) / Berisha, Visar (Committee member) / Arizona State University (Publisher)
Created2018
Description
Human-agent teams (HATs) are expected to play a larger role in future command and control systems where resilience is critical for team effectiveness. The question of how HATs interact to be effective in both normal and unexpected situations is worthy of further examination. Exploratory behaviors are one that way adaptive systems discover opportunities to expand and refine their performance. In this study, team interaction exploration is examined in a HAT composed of a human navigator, human photographer, and a synthetic pilot while they perform a remotely-piloted aerial reconnaissance task. Failures in automation and the synthetic pilot’s autonomy were injected throughout ten missions as roadblocks. Teams were clustered by performance into high-, middle-, and low-performing groups. It was hypothesized that high-performing teams would exchange more text-messages containing unique content or sender-recipient combinations than middle- and low-performing teams, and that teams would exchange less unique messages over time. The results indicate that high-performing teams had more unique team interactions than middle-performing teams. Additionally, teams generally had more exploratory team interactions in the first session of missions than the second session. Implications and suggestions for future work are discussed.
ContributorsLematta, Glenn Joseph (Author) / Chiou, Erin K. (Thesis advisor) / Cooke, Nancy J. (Committee member) / Roscoe, Rod D. (Committee member) / Arizona State University (Publisher)
Created2019
Description
Humans and robots need to work together as a team to accomplish certain shared goals due to the limitations of current robot capabilities. Human assistance is required to accomplish the tasks as human capabilities are often better suited for certain tasks and they complement robot capabilities in many situations. Given the necessity of human-robot teams, it has been long assumed that for the robotic agent to be an effective team member, it must be equipped with automated planning technologies that helps in achieving the goals that have been delegated to it by their human teammates as well as in deducing its own goal to proactively support its human counterpart by inferring their goals. However there has not been any systematic evaluation on the accuracy of this claim.
In my thesis, I perform human factors analysis on effectiveness of such automated planning technologies for remote human-robot teaming. In the first part of my study, I perform an investigation on effectiveness of automated planning in remote human-robot teaming scenarios. In the second part of my study, I perform an investigation on effectiveness of a proactive robot assistant in remote human-robot teaming scenarios.
Both investigations are conducted in a simulated urban search and rescue (USAR) scenario where the human-robot teams are deployed during early phases of an emergency response to explore all areas of the disaster scene. I evaluate through both the studies, how effective is automated planning technology in helping the human-robot teams move closer to human-human teams. I utilize both objective measures (like accuracy and time spent on primary and secondary tasks, Robot Attention Demand, etc.) and a set of subjective Likert-scale questions (on situation awareness, immediacy etc.) to investigate the trade-offs between different types of remote human-robot teams. The results from both the studies seem to suggest that intelligent robots with automated planning capability and proactive support ability is welcomed in general.
In my thesis, I perform human factors analysis on effectiveness of such automated planning technologies for remote human-robot teaming. In the first part of my study, I perform an investigation on effectiveness of automated planning in remote human-robot teaming scenarios. In the second part of my study, I perform an investigation on effectiveness of a proactive robot assistant in remote human-robot teaming scenarios.
Both investigations are conducted in a simulated urban search and rescue (USAR) scenario where the human-robot teams are deployed during early phases of an emergency response to explore all areas of the disaster scene. I evaluate through both the studies, how effective is automated planning technology in helping the human-robot teams move closer to human-human teams. I utilize both objective measures (like accuracy and time spent on primary and secondary tasks, Robot Attention Demand, etc.) and a set of subjective Likert-scale questions (on situation awareness, immediacy etc.) to investigate the trade-offs between different types of remote human-robot teams. The results from both the studies seem to suggest that intelligent robots with automated planning capability and proactive support ability is welcomed in general.
ContributorsNarayanan, Vignesh (Author) / Kambhampati, Subbarao (Thesis advisor) / Zhang, Yu (Thesis advisor) / Cooke, Nancy J. (Committee member) / Fainekos, Georgios (Committee member) / Arizona State University (Publisher)
Created2015
Description
Although current urban search and rescue (USAR) robots are little more than remotely controlled cameras, the end goal is for them to work alongside humans as trusted teammates. Natural language communications and performance data are collected as a team of humans works to carry out a simulated search and rescue task in an uncertain virtual environment. Conditions are tested emulating a remotely controlled robot versus an intelligent one. Differences in performance, situation awareness, trust, workload, and communications are measured. The Intelligent robot condition resulted in higher levels of performance and operator situation awareness (SA).
ContributorsBartlett, Cade Earl (Author) / Cooke, Nancy J. (Thesis advisor) / Kambhampati, Subbarao (Committee member) / Wu, Bing (Committee member) / Arizona State University (Publisher)
Created2015
Description
Speech recognition and keyword detection are becoming increasingly popular applications for mobile systems. While deep neural network (DNN) implementation of these systems have very good performance,
they have large memory and compute resource requirements, making their implementation on a mobile device quite challenging. In this thesis, techniques to reduce the memory and computation cost
of keyword detection and speech recognition networks (or DNNs) are presented.
The first technique is based on representing all weights and biases by a small number of bits and mapping all nodal computations into fixed-point ones with minimal degradation in the
accuracy. Experiments conducted on the Resource Management (RM) database show that for the keyword detection neural network, representing the weights by 5 bits results in a 6 fold reduction in memory compared to a floating point implementation with very little loss in performance. Similarly, for the speech recognition neural network, representing the weights by 6 bits results in a 5 fold reduction in memory while maintaining an error rate similar to a floating point implementation. Additional reduction in memory is achieved by a technique called weight pruning,
where the weights are classified as sensitive and insensitive and the sensitive weights are represented with higher precision. A combination of these two techniques helps reduce the memory
footprint by 81 - 84% for speech recognition and keyword detection networks respectively.
Further reduction in memory size is achieved by judiciously dropping connections for large blocks of weights. The corresponding technique, termed coarse-grain sparsification, introduces
hardware-aware sparsity during DNN training, which leads to efficient weight memory compression and significant reduction in the number of computations during classification without
loss of accuracy. Keyword detection and speech recognition DNNs trained with 75% of the weights dropped and classified with 5-6 bit weight precision effectively reduced the weight memory
requirement by ~95% compared to a fully-connected network with double precision, while showing similar performance in keyword detection accuracy and word error rate.
they have large memory and compute resource requirements, making their implementation on a mobile device quite challenging. In this thesis, techniques to reduce the memory and computation cost
of keyword detection and speech recognition networks (or DNNs) are presented.
The first technique is based on representing all weights and biases by a small number of bits and mapping all nodal computations into fixed-point ones with minimal degradation in the
accuracy. Experiments conducted on the Resource Management (RM) database show that for the keyword detection neural network, representing the weights by 5 bits results in a 6 fold reduction in memory compared to a floating point implementation with very little loss in performance. Similarly, for the speech recognition neural network, representing the weights by 6 bits results in a 5 fold reduction in memory while maintaining an error rate similar to a floating point implementation. Additional reduction in memory is achieved by a technique called weight pruning,
where the weights are classified as sensitive and insensitive and the sensitive weights are represented with higher precision. A combination of these two techniques helps reduce the memory
footprint by 81 - 84% for speech recognition and keyword detection networks respectively.
Further reduction in memory size is achieved by judiciously dropping connections for large blocks of weights. The corresponding technique, termed coarse-grain sparsification, introduces
hardware-aware sparsity during DNN training, which leads to efficient weight memory compression and significant reduction in the number of computations during classification without
loss of accuracy. Keyword detection and speech recognition DNNs trained with 75% of the weights dropped and classified with 5-6 bit weight precision effectively reduced the weight memory
requirement by ~95% compared to a fully-connected network with double precision, while showing similar performance in keyword detection accuracy and word error rate.
ContributorsArunachalam, Sairam (Author) / Chakrabarti, Chaitali (Thesis advisor) / Seo, Jae-Sun (Thesis advisor) / Cao, Yu (Committee member) / Arizona State University (Publisher)
Created2016
Description
Introduction: A diet high in fermented, oligio-, di-, monosaccharide, and polyols
(FODMAP) has been shown to exacerbate symptoms of irritable bowel syndrome
(IBS). Previous literature has shown significant improvement in IBS symptoms after
adherence to a low FODMAP diet (LFD). However, dietary adherence to the LFD is
difficult with patients stating that information provided by healthcare providers
(HCPs) is generalized and nonspecific requiring them to search for supplementary
information to fit their needs. Notably, studies that have used a combination of
online and in-person methods for treatment have shown improved adherence to the
LFD. Objective: To determine whether a novel artificial intelligence (AI) dietary
mobile application will improve adherence to the LFD compared to a standard online
dietary intervention (CON) in populations with IBS or IBS-like symptoms over a 4-
week period. Methods: Participants were randomized into two groups: APP or CON.
The intervention group was provided access to an AI mobile application, a dietary
resource verified by registered dietitians which uses artificial intelligence to
individualize dietary guidance in real-time with the ability to scan menus and
barcodes and provide individuals with food scores based on their dietary preferences.
Primary measures included mobile app engagement, dietary adherence, and
manifestation of IBS-like symptoms. Baseline Results: A total of 58 participants
were randomized to groups. This is an ongoing study and this thesis details the
methodology and baseline characteristics of the participants at baseline and
intervention start. Validation of the application could improve the range of offerings
for lifestyle diseases treatable through dietary modification.
(FODMAP) has been shown to exacerbate symptoms of irritable bowel syndrome
(IBS). Previous literature has shown significant improvement in IBS symptoms after
adherence to a low FODMAP diet (LFD). However, dietary adherence to the LFD is
difficult with patients stating that information provided by healthcare providers
(HCPs) is generalized and nonspecific requiring them to search for supplementary
information to fit their needs. Notably, studies that have used a combination of
online and in-person methods for treatment have shown improved adherence to the
LFD. Objective: To determine whether a novel artificial intelligence (AI) dietary
mobile application will improve adherence to the LFD compared to a standard online
dietary intervention (CON) in populations with IBS or IBS-like symptoms over a 4-
week period. Methods: Participants were randomized into two groups: APP or CON.
The intervention group was provided access to an AI mobile application, a dietary
resource verified by registered dietitians which uses artificial intelligence to
individualize dietary guidance in real-time with the ability to scan menus and
barcodes and provide individuals with food scores based on their dietary preferences.
Primary measures included mobile app engagement, dietary adherence, and
manifestation of IBS-like symptoms. Baseline Results: A total of 58 participants
were randomized to groups. This is an ongoing study and this thesis details the
methodology and baseline characteristics of the participants at baseline and
intervention start. Validation of the application could improve the range of offerings
for lifestyle diseases treatable through dietary modification.
ContributorsRafferty, Aaron (Author) / Johnston, Carol (Thesis advisor) / Hall, Richard (Committee member) / Fitton, Renee (Committee member) / Arizona State University (Publisher)
Created2020