Thomas Baldi, Giovanni Delnevo, Roberto Girau, S. Mirri
Environmental conditions within vehicles represent a significant element of the driver's well-being and comfort. In particular, exposure to air pollution has been proven to affect human cognitive performances, hence it could represent a risk to driving safety. Monitoring internal and external environmental data could provide interesting hints, helpful in predicting trends and situations potentially dangerous and/or unease, that should be reported, enhancing the driver's awareness. This paper presents a study we have conducted with the aim of predicting indoor vehicle environmental conditions, thanks to a campaign of data collection. In particular, we have adopted a multi-sensor kit, installed within and outside a vehicle, then we have exploited driving sessions in a urban environment. Different machine learning algorithms have been adopted to test their accuracy in predicting internal conditions, on the basis of external ones, discussing the obtained results.
{"title":"On the prediction of air quality within vehicles using outdoor air pollution: sensors and machine learning algorithms","authors":"Thomas Baldi, Giovanni Delnevo, Roberto Girau, S. Mirri","doi":"10.1145/3538393.3544934","DOIUrl":"https://doi.org/10.1145/3538393.3544934","url":null,"abstract":"Environmental conditions within vehicles represent a significant element of the driver's well-being and comfort. In particular, exposure to air pollution has been proven to affect human cognitive performances, hence it could represent a risk to driving safety. Monitoring internal and external environmental data could provide interesting hints, helpful in predicting trends and situations potentially dangerous and/or unease, that should be reported, enhancing the driver's awareness. This paper presents a study we have conducted with the aim of predicting indoor vehicle environmental conditions, thanks to a campaign of data collection. In particular, we have adopted a multi-sensor kit, installed within and outside a vehicle, then we have exploited driving sessions in a urban environment. Different machine learning algorithms have been adopted to test their accuracy in predicting internal conditions, on the basis of external ones, discussing the obtained results.","PeriodicalId":438536,"journal":{"name":"Proceedings of the ACM SIGCOMM Workshop on Networked Sensing Systems for a Sustainable Society","volume":"18 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124033357","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Udhaya Kumar Dayalan, Rostand A. K. Fezeu, T. Salo, Zhi-Li Zhang
We introduce Kaala, a scalable, hybrid, end-to-end IoT system simulator that can integrate with diverse, real-world IoT cloud services. Many IoT simulators run in isolation and do not interface with real-world IoT cloud systems or servers. This isolation makes it difficult for experiments to fully replicate the diversity that exists in end-to-end, real-world systems. Kaala is intended to bridge the gap between IoT simulation experiments and the real world. The simulator can interact with cloud IoT services, such as those offered by Amazon, Microsoft and Google. Kaala leverages vendor-provided software development kits (SDKs) to implement the vendor-specific protocols that are necessary permit simulated IoT devices and gateways to seamlessly communicate with real-world cloud IoT systems. Kaala has the ability to simulate a large number of diverse IoT devices, as well as to simulate events that may simultaneously affect several sensors. Evaluation results show that Kaala is able to, with minimal overhead, seamlessly connect simulated IoT devices to real-world cloud IoT systems.
{"title":"Kaala","authors":"Udhaya Kumar Dayalan, Rostand A. K. Fezeu, T. Salo, Zhi-Li Zhang","doi":"10.1145/3538393.3544937","DOIUrl":"https://doi.org/10.1145/3538393.3544937","url":null,"abstract":"We introduce Kaala, a scalable, hybrid, end-to-end IoT system simulator that can integrate with diverse, real-world IoT cloud services. Many IoT simulators run in isolation and do not interface with real-world IoT cloud systems or servers. This isolation makes it difficult for experiments to fully replicate the diversity that exists in end-to-end, real-world systems. Kaala is intended to bridge the gap between IoT simulation experiments and the real world. The simulator can interact with cloud IoT services, such as those offered by Amazon, Microsoft and Google. Kaala leverages vendor-provided software development kits (SDKs) to implement the vendor-specific protocols that are necessary permit simulated IoT devices and gateways to seamlessly communicate with real-world cloud IoT systems. Kaala has the ability to simulate a large number of diverse IoT devices, as well as to simulate events that may simultaneously affect several sensors. Evaluation results show that Kaala is able to, with minimal overhead, seamlessly connect simulated IoT devices to real-world cloud IoT systems.","PeriodicalId":438536,"journal":{"name":"Proceedings of the ACM SIGCOMM Workshop on Networked Sensing Systems for a Sustainable Society","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124800440","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Although Deep Learning (DL) is revolutionising practices across fields, it requires a large amount of data and computing resources, requires considerable training time, and is thus expensive. This study proposes a transfer learning approach by adopting a simplified version of a standard Convolution Neural Network (CNN), which is successful in another domain. We explored three transfer learning approaches: freezing all layers except the first and the last layer of the CNN model, which we had modified, freezing the first layer, updating the weights of the rest of the layers, and fine-tuning the entire network. Furthermore, we trained a DL model from scratch to act as a baseline. We performed the experiments on the Edge Impulse platform. We evaluated the models based on plant-village, tea diseases and land use datasets. Fine-tuning and training the whole network produced the best precision, accuracy, recall, f-measure and sensitivity across the datasets. All three transfer learning schemes significantly reduced the training by more than half. Further, we deployed the fine-tuned model in detecting diseases in tea two months after the idea's conception, and it showed a good correlation with the experts' decisions. The evaluation results showed that it is viable to perform transfer learning among domains to accelerate solutions deployments. Additionally, Edge Impulse is ideal in resource-constrained environments, especially in developing countries lacking computing resources and expertise to train DL models from scratch. This insight can propel the development and rollout of various applications addressing the Sustainable Development Goals targeted at zero hunger and no poverty, among other goals.
{"title":"An instance-based deep transfer learning approach for resource-constrained environments","authors":"Gibson Kimutai, Anna Förster","doi":"10.1145/3538393.3544938","DOIUrl":"https://doi.org/10.1145/3538393.3544938","url":null,"abstract":"Although Deep Learning (DL) is revolutionising practices across fields, it requires a large amount of data and computing resources, requires considerable training time, and is thus expensive. This study proposes a transfer learning approach by adopting a simplified version of a standard Convolution Neural Network (CNN), which is successful in another domain. We explored three transfer learning approaches: freezing all layers except the first and the last layer of the CNN model, which we had modified, freezing the first layer, updating the weights of the rest of the layers, and fine-tuning the entire network. Furthermore, we trained a DL model from scratch to act as a baseline. We performed the experiments on the Edge Impulse platform. We evaluated the models based on plant-village, tea diseases and land use datasets. Fine-tuning and training the whole network produced the best precision, accuracy, recall, f-measure and sensitivity across the datasets. All three transfer learning schemes significantly reduced the training by more than half. Further, we deployed the fine-tuned model in detecting diseases in tea two months after the idea's conception, and it showed a good correlation with the experts' decisions. The evaluation results showed that it is viable to perform transfer learning among domains to accelerate solutions deployments. Additionally, Edge Impulse is ideal in resource-constrained environments, especially in developing countries lacking computing resources and expertise to train DL models from scratch. This insight can propel the development and rollout of various applications addressing the Sustainable Development Goals targeted at zero hunger and no poverty, among other goals.","PeriodicalId":438536,"journal":{"name":"Proceedings of the ACM SIGCOMM Workshop on Networked Sensing Systems for a Sustainable Society","volume":"61 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131323134","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Poor air quality has been responsible for millions of premature deaths. Acknowledging the critical role air quality plays in the future of their populations, governments across the world have been installing networks of fixed location air quality measurement instruments. But these monitoring stations are expensive and therefore spatially sparse, typically publishing summaries of hourly averages of pollutant measurements once per day. Data so sparse spatially and temporally offers little to inform the street user or policy maker as to what is happening at a more granular level, thus reducing the ability to avoid pollutants. This paper investigates the feasibility of using consumer grade mobile sensors as a means to contribute to a real time federated hyper-local crowd sensing air quality data service, RealTimeAir (RTA), underpinned by government reference sensors. We compare two mobile sensors and examine the correlation of the measurements between them. We investigate the correlation between these sensors and the more expensive fixed monitoring stations. We consider the variation of measurements over time and space to investigate the need for greater granularity of these measurements. Finally, we present a low pollutant exposure route finder as a use case for the proposed system.
{"title":"RealTimeAir","authors":"S. Hart, Joseph Doyle","doi":"10.1145/3538393.3544933","DOIUrl":"https://doi.org/10.1145/3538393.3544933","url":null,"abstract":"Poor air quality has been responsible for millions of premature deaths. Acknowledging the critical role air quality plays in the future of their populations, governments across the world have been installing networks of fixed location air quality measurement instruments. But these monitoring stations are expensive and therefore spatially sparse, typically publishing summaries of hourly averages of pollutant measurements once per day. Data so sparse spatially and temporally offers little to inform the street user or policy maker as to what is happening at a more granular level, thus reducing the ability to avoid pollutants. This paper investigates the feasibility of using consumer grade mobile sensors as a means to contribute to a real time federated hyper-local crowd sensing air quality data service, RealTimeAir (RTA), underpinned by government reference sensors. We compare two mobile sensors and examine the correlation of the measurements between them. We investigate the correlation between these sensors and the more expensive fixed monitoring stations. We consider the variation of measurements over time and space to investigate the need for greater granularity of these measurements. Finally, we present a low pollutant exposure route finder as a use case for the proposed system.","PeriodicalId":438536,"journal":{"name":"Proceedings of the ACM SIGCOMM Workshop on Networked Sensing Systems for a Sustainable Society","volume":"83 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124599132","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
This paper addresses a wireless sensor network technology that supports the deployment of sustainable IoT applications essential to future zero-carbon smart cities. We propose a novel data collection and retrieval scheme to adopt an information-centric network into wireless sensor networks for energy efficiency. The results of laboratory-based experiments using a testbed and prototype network demonstrate the feasibility and applicability of the proposed scheme in terms of network throughput, latency, jitter, and energy consumption.
{"title":"A preliminary analysis of data collection and retrieval scheme for green information-centric wireless sensor networks","authors":"Shintaro Mori","doi":"10.1145/3538393.3544932","DOIUrl":"https://doi.org/10.1145/3538393.3544932","url":null,"abstract":"This paper addresses a wireless sensor network technology that supports the deployment of sustainable IoT applications essential to future zero-carbon smart cities. We propose a novel data collection and retrieval scheme to adopt an information-centric network into wireless sensor networks for energy efficiency. The results of laboratory-based experiments using a testbed and prototype network demonstrate the feasibility and applicability of the proposed scheme in terms of network throughput, latency, jitter, and energy consumption.","PeriodicalId":438536,"journal":{"name":"Proceedings of the ACM SIGCOMM Workshop on Networked Sensing Systems for a Sustainable Society","volume":"248 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122068740","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Edge computing is a network architecture in which computing and storage capabilities are moved at the fringes of the Internet, close to the end-users. The main goal of edge computing is to enable responsive services, thanks to much shorter paths compared to the ones encountered when communicating with remotely positioned cloud servers. In this paper, we report experimental results concerning an overlooked benefit of edge computing: energy is saved on client devices. We carried out an experimental evaluation using both software-based and hardware-based energy estimation methods. Results show that, for HTTP-based communication, the lifetime of a device can be extended significantly when using the edge instead of a remote cloud.
{"title":"Saving energy on smartphones through edge computing: an experimental evaluation","authors":"Chiara Caiazza, Valerio Luconi, Alessio Vecchio","doi":"10.1145/3538393.3544935","DOIUrl":"https://doi.org/10.1145/3538393.3544935","url":null,"abstract":"Edge computing is a network architecture in which computing and storage capabilities are moved at the fringes of the Internet, close to the end-users. The main goal of edge computing is to enable responsive services, thanks to much shorter paths compared to the ones encountered when communicating with remotely positioned cloud servers. In this paper, we report experimental results concerning an overlooked benefit of edge computing: energy is saved on client devices. We carried out an experimental evaluation using both software-based and hardware-based energy estimation methods. Results show that, for HTTP-based communication, the lifetime of a device can be extended significantly when using the edge instead of a remote cloud.","PeriodicalId":438536,"journal":{"name":"Proceedings of the ACM SIGCOMM Workshop on Networked Sensing Systems for a Sustainable Society","volume":"38 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115209885","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
IoT devices are becoming increasingly complex, support multiple sensors and often rely on batteries and renewable energy. Scheduling algorithms can help to manage their energy usage. When multiple devices cooperatively monitor an environment, scheduling sensing tasks across a distributed set of IoT devices can be challenging because they have limited information about other devices, limited energy and communication bandwidth. In addition, sharing information between devices can be costly in terms of energy. Our Tier-based Task scheduling protocol (T2), is an energy efficient distributed scheduler for a network of multi-sensor IoT devices. T2, adapting on an epoch-by-epoch basis distributes task executions throughout an epoch to minimize temporal sensing overlap without exceeding task deadlines. Our experiments show that T2 schedules an IoT device's sensing task start time before its deadline expires. When compared against a simple periodic scheduler, T2 schedules closer to the optimal centralized EDF scheduler.
{"title":"A renewable energy-aware distributed task scheduler for multi-sensor IoT networks","authors":"Elizabeth Liri, K. Ramakrishnan, K. Kar","doi":"10.1145/3538393.3544936","DOIUrl":"https://doi.org/10.1145/3538393.3544936","url":null,"abstract":"IoT devices are becoming increasingly complex, support multiple sensors and often rely on batteries and renewable energy. Scheduling algorithms can help to manage their energy usage. When multiple devices cooperatively monitor an environment, scheduling sensing tasks across a distributed set of IoT devices can be challenging because they have limited information about other devices, limited energy and communication bandwidth. In addition, sharing information between devices can be costly in terms of energy. Our Tier-based Task scheduling protocol (T2), is an energy efficient distributed scheduler for a network of multi-sensor IoT devices. T2, adapting on an epoch-by-epoch basis distributes task executions throughout an epoch to minimize temporal sensing overlap without exceeding task deadlines. Our experiments show that T2 schedules an IoT device's sensing task start time before its deadline expires. When compared against a simple periodic scheduler, T2 schedules closer to the optimal centralized EDF scheduler.","PeriodicalId":438536,"journal":{"name":"Proceedings of the ACM SIGCOMM Workshop on Networked Sensing Systems for a Sustainable Society","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127723642","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：