(Workshop) Tiny Edge Machine Learning: Challenges and Opportunities for IoT Applications

(Workshop) Tiny Edge Machine Learning: Challenges and Opportunities for IoT Applications

Show Video

uh we'll continue with us uh when he start uh when he is in uh um yes so this is the the the workshop that you registered for or invited to uh which is the tiny edge machine learning challenges and opportunities for iot applications and thank you to each one of you for being here with us today we are very pleased to be able to welcome those you that have been with us for a long time now as well as those uh who are new to the hashimoto university and the royal academy of engineering so today is the first online workshop and meeting for our funded project entitled the ultra low cost machine learning enabled iot devices on the edge and we are so proud here in the hashemite university and the faculty of engineering to be able to host it today in this zoo online meeting so the aim of the workshop is as the aim of the window of distinguished international associate projects is to spotlight the capabilities of the low performance edge sensor node controllers by using tiny machine learning algorithms in order to reduce the expected congestion and capital expenditure of the iud infrastructure especially in the agriculture applications in the cost in the context of smart farming and water management in jordan we believe that this will ensure that iot agriculture is inclusive not to reach only high-tech farming industry but also the small and medium-sized pool farmers this workshop is a kick-off the workshop mainly with some technical talks on tiny machine learning and from many countries jordan uk italy and the gulf uh hopefully the speaker from italy will be in with us he sent me that he has some technical problems and we we hope that he will be with us this is a project that i'm leading from the athena might university and is funded by the royal academy of engineering in the uk and also funded by the hashimoto university under the title ultra low cost machine learning the workshop will highlight the state-of-the-art technologies and tiny machine learning the funding windows by the royal academy of engineering as will be presented by mr shane and matthew from the royal academy of engineering the head of international partnership and collaborations in the academy it will highlight also the collaboration opportunities that may raise the efficiency of students in energy and engineering faculty and information technology through advanced research graduation projects and to be and to provide them with skills that qualify them to in the labor markets through a series of workshops such like this and courses in addition to networking with number of international universities research institutions and local market companies chances media have all chosen us to be a part of this workshop due to mutual passion for us for our universities companies and the academy your passions help us also tonight gather and the energy of this workshop will create what will allow us to achieve our individuals and targets as well as the collaboration goals during the next few months we will be learning via several workshops seminars special events and conferences through our planned activities in this project where you will be able to join in and get hands-on experience i hope that these will help you who are interested in growing their knowledge and strengthening the cooperation through this kind of networking events at this stage i would like to pass my a grateful thanks to the royal academy of engineering for the support and the funding and believing in us here in the faculty of engineering at the hashimoto university through a similar civil funding windows that funds multiple projects in the hashimoto university mainly namely the distinguished international associate window that this project is funded by and the other one which is transforming systems through partnership the tsp which is known in in jordan here as a new templating window so a very warm uh welcome to each and every one of you if you have any question suggestion clarification or just want to see how you can raise your hand and ask your question or send us uh on the chat well um i don't want to take too much of your time and i need to leave some time for the professor for wazada the president of the hashimoto university who is always a stronger strong believer and supporter to the research efforts that we are doing and who is looking always towards strengthening the collaboration with the different entities and i think this workshop is a good start at the hashimoto university with the royal academy of engineering to build to build upon our previous school operations and job into more and more strong and direct collaborations uh dr for was the mic is yours i think it's muted okay yeah yeah can you hear me yes thank you very much indeed it's my pleasure to be with you i am not an engineer but i am a linguistic engineer we have so much in common between linguistics and engineering uh from day one when i was assuming the uh leadership of this university uh ali hayajna in particular and the faculty of engineering have won my admiration they are great guys they did they are leading university so the faculty of engineering for me is really indeed a milestone a landmark that are achieving uh something that will lead the university from globality from locality to globality so and second i was so happy to see many faces who are gathering with us attending including my vice president professor dr harry who is my vice president for that faculty and he is coming from the faculty of engineering and just to let you know more than 10 people in administration are coming from engineering there is no wonder therefore we are having a kind of envy from other faculties why engineering in particular engineering are not only for engineering purposes they are engineering the generation i think we need this uh a new definition let me go back to what has been written to me to be said in your committee i'm really happy to be with you in this workshop which is organized by the college of engineering and entitled tiny edge machine learning challenges and opportunities for iot application and i mentioned ali by his name when i met his majesty king abdullah in the 19th of april 2020 when we were talking about his uh project and the imp the application of that project in [Music] the arid land the role of using drones developing certain specific technology for and this is actually the tiny coming from this so and also i would like to welcome you all the attendees from all over the world from inside jordan and outside jordan moreover i welcome the partners and event speakers participating in the pioneering project supported by the royal company and i will pose here they are royal committee academy and we are hashimoto are royal too so this is one of the common elements between the two royals the royal academy and hashimai university and the royal engineering faculty at hashimoto university thank you for the support and funding provided by the royal academy through the various programs for a number of projects from the hashimoto university and in various jordanian universities as you know the hashimoto university is oriented toward achieving an academic pioneering position and excellence in university teaching scientific research at both national regional and to serve society through its educational functions and to participate in achieving the goals of the comprehensive national development through preparing loyal and dedicated men and women who are not only technically competent in their professional fields but also lifelong learners who have a breadth vision loyalty to their nation and the sense of civic and moral responsibility and devotion to the fundamental values of human life we are also interested in the academy partnership with the jordanian higher education sector to enhance and develop capabilities within jordan and higher education and research institutions and to enhance the output in education and research and engineering innovation by strengthening industrial links and benefiting from the uk expertise especially in every important areas of the global local such as artificial intelligence and internet of things and i hope that this is going to be a department someday sometime either within the faculty of technology or the faculty of engineering and here i would like to point out that we are interested in these scientific disciplines and their application to serve all productive and industrial sectors we have recently developed an academic program in artificial intelligence and we have centers at the university concerned with innovation excellence big data and artificial intelligence and we seek to develop all of these through the projects last but not least i extend my thanks to the royal academy for the unlimited support for their projects and ideas of faculty members of the hashimoto university and we look forward to further cooperation i also thank colleagues from the college of edu engineering for organizing this workshop and for their creativity in this pioneering project we promise that we all continue to provide absolute support for this collaboration thank you very much indeed for listening and wish you all the best thank you okay thank you professor and thank you for all of your participants uh here in the workshop uh as the schedule told us the time is now for mr uh shane matthew from the royal academy of engineering to give his talk mr shane the mic is yours um thank you very much uh dr ali and um i'd love to thank you and uh professor follows for an excellent redux introduction and um for the kind words you've said about the academy's international work and i'd like uh to take a few minutes to talk about um what the economy does internationally and focusing particularly on some of the opportunities um dr ali haya jenna um mentioned earlier on about um that are available to researchers from jordan and i'd like to share my screen is that that's okay um dr ali you as a whole oh yeah okay thank you let me do that now thanks this will show stuff other screen sharing yes if that's okay isn't that right i can then just that's fine okay sure so can you see a presentation now yes has that worked dr ali not yet yes yes it's okay it's working now yes yes okay thank you um so yeah i'll just do a short give a short presentation on our international activities and if i move on so the academy's mission global mission is to advance engineers contribution to a safer inclusive more sustainable world with a particular emphasis on tackling the global challenges of our time and we have a really grown global portfolio of programs and with a lot of focus on developing and emerging countries uh we have recently adopted a strategy to 2025 that puts the global work at the heart of what the academy does and its goal is to harness the power of engineering for a sustainable global society and an inclusive uk economy and there's three things we do that really help address those those goals fostering talent and diversity uh promoting innovation and influencing policy in the public and i'll go quickly through what we do uh that meets each of these goals but obviously we can only re meet these goals with key partners around the world and i'd like to pay tribute to the jordanian ministry and particularly to the hashemite university for being such excellent partners i'll talk very quickly about our work innovation uh we have two um real um goals here uh the first is to um stimulate uh better so more effective international innovation collaborations so not just to accelerate development but also um to ensure that to accelerate the deployment of tools to advance sustainability and we also uh want to increase our support for excellent researchers and entrepreneurs uh in across the world developing innovations targeted sustainability challenges and how we do those things um so the transforming systems through partnership program and that dr riley spoke about earlier is our main vehicle for um stimulating more effective international innovation collaborations and we work in partnership with jordan and six other countries around the world in this and really it's a mechanism um to allow uh a university in the um the the host country to catalyze a partnership of um not just academia but industry government stakeholders and smes and civil society stakeholders and mobilize them around a national challenge and this uh this program at the moment we are in the uk we're going through uh um uh a three-year spending review uh across the whole of the country um and as you may have seen uh this program was previously funded through um our our aid development funding which um has reduced because of the impact of the covet pandemic in the uk however we are um we are putting uh bids to for government to expand this program under other streams of money and hopefully we will have good news about future rounds of this scheme um later in the year or early in 2022. we also provide support for researchers and entrepreneurs developing sustainability innovations and that's another area where we partner currently with jordan and what's called the leaders in innovation fellowships so this is really an innovation and entrepreneurship boot camp for um for engineering researchers with an innovation that tackles and a sustainable development goal that they're trying to commercialize and it's um it's a very intensive boot camp and at the end of it along with uh you know a network in the uk and a slide deck um because there's lots of um training on um uh both uh running uh startup and how to um how to um ensure finance for the startup the the other thing we want to see is a mindset change spreading around academia and so that um across engineered researchers and there's an understanding that entrepreneurship is something that adds value and is not a kind of a side alley it's an important part of an engineering research career we also run a couple of africa focused and sub-saharan african-focused programs the africa prize for engineering innovation and project care which was a programme uh to help entrepreneurs in sub-saharan africa during the pandemic um i'll mention one africa prize winner last year's africa prize winner because uh her work is very related to the topic of this conference this is uh charlotte negesian from cote d'ivoire and um we're delighted to support her her um innovation called base api was a facial identifying technology so that so it allowed um finance companies in sub-saharan africa um to verify identities through facial recognition through essentially a camera phone uh on the person's phone and it was um a used ai but it was specifically geared towards recognizing african faces and it's an example i think of low-cost and machine learning interventions in low resource environments that really add value uh talk very quickly about our policy engagement work uh again the um the work we do is generally a partnership with engineering academies around the world and with global engineering stakeholders and it's about embedding a systems approach in responses to sustainability challenges and developing strategic alliances of engineering stakeholders to inform and engage policy makers and really improve global engineering leadership on the challenges that matters to the engineering community and uh i'll very briefly talk about this uh we run three missions with globally um which are um active around the world it's not in specific partnerships with specific countries around better management of safer complex systems improving the end of engineered life of um of of of engineered objects especially we're focusing at the moment on um large-scale object ships and maritime infrastructures but we have also done a review of a global waste streams and we're carrying out a piece of work for cup 27 in africa on the challenge of open burning and how open burning of waste contributes and not just to health and safety problems but also contributes massively to global and co2 emissions and then pandemic preparedness uh helping engineers through the with innovations that um address the pandemic and assessing how the engineering community performed and we also have partnerships uh specifically with um a number of countries on a number of themes uh an example of the um the the safer end of engineered life work um global global shipping um 90 of global shifts are decommissioned uh in a very informal way on two or three beaches in india and bangladesh and uh what we are trying to do is work with stakeholders within the industry within manufacturing and also within those countries uh to improve the the the safety of the process and then i'm going to say a little more about thompson diversity because that is uh where the distinguished um international associates program that dr ali is part of and sits and that is now open for um there's no open for application so i'll tell you a little bit more about that again our goals here are to mobilize the global engineering community create strong international alliances support sustainability and support capacity building and engineering as a priority and we do a lot of that in sub-saharan africa and then um there's also a a uk element ensuring that the uk workforce is world leading and truly inclusive uh we do the world leading and truly inclusive workforce through a global challenge visa scheme which allows engineers and who are emerging talents or global talents um from anywhere in the world uh to come to the uk with a pathway and through to within three years and to settlement uh and also the distinguished international associates program which i'll talk about in a second uh we support capacity building and mainly in sub-saharan africa through programs addressing higher education and institutions building their links with industry and through a program called africa catalyst which supports professional engineering institutions and a new program on engineering skills where they're most needed which uh tries to ensure that um all countries in sub-saharan africa have the engineering skills at their safety skills at their disposal uh to safely install uh maintain and repair uh critical infrastructure and we also run a number of programs bringing researchers from all countries together and from all this engineering disciplines together around a common challenge and they're called frontiers programs we run a bigger version of this with the us and china called global ground challenges someone's that's more focused on students finally i'll just say a few words if i may about the distinguished international associates program and that dr ali is is one of our treasured associates this year it's a new program and it is open to excellent international engineers working across all sectors at the cutting edge of engineering research or innovation who already have collaborations or connections to the uk but they want to iden intensify those and what it aims to do is really support and strengthen collaborations and engagement between awardees and their peers in the uk engineering community and it's particularly broadening out those links from maybe one or two collaborators to a broader range of uk universities and engineers and we also provide funding communication support and to empower the associates to broaden and demonstrate their links with the uk and the academy and we promote their activities particularly where they align with those academy objectives of sustainability for all as this conference very very clearly does and we also hope to build a peer support p a peer support network of these associates and to help and stimulate the engineering communities and the home countries uh and facilitate opportunities for the alumni to build their own links and be it on research or innovation or policy with the whole of the uk's engineering community so there's funding of 10 000 pounds in the uk which is about 12 000 euros um for up to 24 months to strengthen collaborations with the uk and you're um we are starting up an associate's network and so it allows participation in that but also you get tailored support and engaging with the academy to pursue your your project um so round two is now open round one it was a pilot round and it was only open to people who had a previous association with the academy as dr ali did through a transforming systems true partnership project this round is open to any engineer with any excellent engineer with strong links with the uk and um they will join our first cohort in a growing international network uh the deadline for applications is the 8th of november 2021 and there are links on the presentation which i can ask dr ali to share around and on our website and a um team email for any questions and thank you very much for listening i'm now going to stop sharing and hand back over to dr ali thank you mr shane for uh being here with us today and uh uh delivering your talk and we are very happy here in the hashimoto university to be part of partnerships with the caring for engineering since we have some time i'd like to also as scheduled uh put the claw for a professor for professor owning the dean of the faculty of engineering in the ashmart university and if we we have some time for a short discussion with you it will be after uh dr audi's talk doctor audi the mike is yours thank you so much dr ali for this opportunity in fact many thanks to the president uh professor uh our president professor fawaz abdel haq for his support and the attendance of this workshop this workshop is uh aligned with his vision of supporting every effort to cooperate globally as well as a big thanks in fact to our colleague and vice president uh professor khaled who is a specialist in fact in engineering and construction management and one of his research interests is the artificial intelligence and he used to deal with uh since 20 years with the genetic algorithms and neural networks and this is why he is very interested in in this in this workshop and many thanks in fact to mr shem from the royal academy of engineering for being with us and for his comprehensive talk about the activities and programs of the royal academy so thank you very much for this opportunity uh to talk to the participat participating friends and speakers in this important workshop which deal with a very important topic namely the artificial intelligence and internet things and their applications as well it's my honor in fact to be with you all i would like to welcome all attendees and partners from all over the world and it is real opportunity to thank and appreciate the partnership with the royal academy of engineering in uk we are proud in fact as dr ali mentioned that in the college of engineering we have number of projects funded by by the royal academy we fully believe that holding such specialized workshops and such partnerships aligned with the vision and mission of our faculty of engineering in which we aim to prepare highly qualified graduates for careers in the engineering profession providing them with approach stimulating engineering education professional skills and knowledge that will enable them to succeed in their future career and to prepare graduates to engage in work environment handling different challenges of local regional and global marketplace and also in our high uh post-graduate program to programs to conduct research that is to recognize scientific contribution in applied field globally and looking which support the comprehensive sustainable national plans in in jordan the faculty of engineering is committed to provide students with up-to-date knowledge and practices how are advancing jordan and the region in the field of engineering technology we also committed to research that leads to better solutions for in the industrial uh sector in engineering related problems there is no doubt that artificial intelligence and the internet things are among the most important skills and knowledge that our students must be able to use to raise their competitiveness and employability in the labor market in all of our engineering programs and we have in fact eight engineering departments or programs one of the main educational objectives are to produce graduates who can apply breadth and depth knowledge critical thinking and problem solving skills to excel in official interiors and address contemporary issues related to the engineering discipline the outcome of this project which is lead by dr ali and this workshop are fully compatible with this main objective of the engineering faculty moreover this project and the knowledge of ai and iot are crucial to be employed in the hashemite university where we adopted a comprehensive strategy for environmental management which is directed to a green canvas the strategy is a multi-spectrum of practices enabled and implemented throughout the the campus in the hashemite industry this strategy can be summarized as the following renewable energy energy energy efficiency water management pollution mitigation and capacity building these practices were realized into the fabric of the campus in all infrastructure projects and expansions along with all social and academic procedures it is necessary to emphasize that the faculty of engineering play a crucial role in achieving this vision we are very interested and supported by we are very supported by the university administration represented by uh our president professor fawaz abdelhad and we are very interested in the fourth industrial revolution and hence in fact next week under his uh patronage uh under the patronage of victor rafawz we will we will sign three mous with the most important jordanian companies and institutions working in the field of artificial intel intelligence internet films and robotics and in fact this project in which we are attending this workshop as a part of it will integrate our integral undergraduate students through the graduate projects and it is to be mentioned here as dr ali told me that 15 of our graduate students they are already work under the title of this project and we have two technical workshops have been held and three indirect international scholarships and their posts seem to be granted to students to complete their studies in international universities with direct coordination by dr ali hayajan and the funding institute all under the benefit of this field this project will definitely contribute in transforming knowledge between uk and the our jordanian university through strong and efficient collaboration this policy is linking in fact universities from jordan from uk italy usa vietnam and other companies in jordan one of them is larimar and the way for innovate and is to be mentioned here that the founder of larimar company uh she is a engineer abira vashiti and she is one of our graduates and in fact we we are proud of engineer finally i would like to thank the royal academy of engineering and the representative mr shim he and who's the head of the international partnerships for being with us today and in fact uh i have to mention that uh uh mr shain that we are looking forward for the opportunity opportunity for direct collaboration agreements with the royal academy to fund the projects by researchers from the hashemite university and from the faculty of engineering we we are eager to this opportunity and for uh this collaboration with a direct impact collaboration with the uh with the royal academy thank you so much thank you dr ali and florida zeus thank you thank you professor auni for your talk and thank you all for keeping listening with us in this workshop uh i i think we have a couple of minutes for any discussion if if anyone have if any one of you has any question any clarification or just want to say hi then you can unmute yourself and just just talk whatever you want if you all know me just a couple of minutes yes you evolved and welcome i wanted to ask you to to intervene in in this workshop but uh the mic is yours right you know with your validity uh first of all i would like to thank our president uh professor for attending this workshop and i want to thank you dr ali for taking the lead in organizing this workshop i would like that i want just to strengthen importance of the topic of this uh workshop we are looking for uh innovative research in the internet of things and i would like to just to call for an interdisciplinary research focusing on artificial intelligence applications or real things for and also for internet of things thank you again for taking taking the lead in organizing this workshop and we are looking forward for more workshops in the future thank you that's fine thank you dr khalid for this and i'm i'm i have a plan to speak about the multidisciplinary applications for artificial intelligence and machine learning especially on the edge because when we when we say the edge we say ultra low cost and ultra low performance capabilities of the controllers and the sensor nodes so employing or harnessing these kind of edge devices is very crucial and important to deploy and develop a commercial products which is always the uh the target and the motivation of the engineering especially when we are talking about the hashimoto university which is one of the leading universities in the engineering [Music] this goal is more than uh welcomed and uh we have another we have uh in our activity plan another workshop and we hope so that this workshop will be a physical uh workshop in amman or in in the hashimoto university itself uh i will send you uh an invitations once it's uh settled and uh organized it will be i think and next may the next year so uh i would like to thank you also again uh for being uh with us another uh couple of minutes for anyone who wants to ask any question uh since mr shane and dr awning and dr khalid is with us today i think uh any question would pick its direct orientation to to the managers and the leaders in the university and the academy of engineering you can mute yourself and ask any question directly um i've just wrote my question in the chat yeah i asked will this collaboration take effect in the future and uh was there any project achieved or is there any project currently in progress between these collaborating institutions yeah um okay so as i mentioned before this uh and on the website of the project and the web and workshop uh this project and workshop is part of uh two workshops two projects this project is through the distinguished international associate window uh under the name of ultra low cost low performance edge devices and it's it's it's currently uh going on and and we still have uh around 18 months in this project there is another project through the newton hardy project uh project windows and for drone assisted micro irrigation and and when we talk about uh artificial intelligence and machine learning it takes its part uh in in these projects very strongly so the collaboration is always open you can just send me an email or send me a message and then i will organize for a meeting and physical meeting or online meeting or a special event to talk about any further and future collaboration this workshop is part of the as i i said before part of the window of distinguished international associate but without being funded via the newton kali project or the transforming system for partnership projects then this the ia will not be funded because to be funded with a distinguished international social project you need to be a alumni of the royal academy of engineering and i we're uh one of the aluminized through the newton collegiate project so the newton county project i think will be opened in in a couple of months uh if there is if things goes as usual and uh the call is open for anyone who is listening or he will listen to this recorded meeting to contact us and to collaborate this is my answer to this and and thank you for being here with us so i've been waiting for uh mr dr massimo from italy from the university of mediterra of calpia i'm not sure that he is in with us he told me that he he has some technical issues and circumstances i will put right now dr syed alireza from leeds unit okay shane you you want to oh sorry just um thank you very much and dr ali for your answer to that just to just to say two things in addition that was a very good uh roundup of the the projects that are on offer and two things in addition um the distinguished international associates from this year you do not need to be an alumni of the academy that was just of academy programs that was just the case in the first year which was a pilot year so the current round is open to any excellent researcher with a project um who have already have connections or a project with the uk that they want to intensify and and and then secondly just to say obviously in the uk there's a big spending review in progress and we won't know until it finishes what the um the the outcomes for things like the newton fund will be and we have obviously submitted all our bids uh and we hope even if it doesn't go ahead in the current form that we will be able to have some kind of transforming systems through partnership um uh future future call it might not be under the newton fund or under the aid budget and we will definitely actually know as soon as we find out but maybe there um there may be a pause in calls early next year um while while the result of the spending review filter through all right okay thank you and um i'm very uh pleasant to to hear that the distinguished industry associate is is open for uh any uh researcher who is in uh who is fortunate to to do uh any collaboration with you um from uh honey can we elaborate more about what we mean when we say the projects yeah i can i can do that and it's the broadest possible definition of that so some research collaboration with the uk either existing or previous that you want to um build upon we would also accept for example someone who did their phd in the uk and restraint retained strong links with the university and that department they worked with and wants to start a new project that will kind of build on those links um so it's it's not so the the criteria are not so much about um they're about the purpose of the collaboration uh not about the existing strength yes yeah so thank you shane for this and yeah even if you are a phd student masters student in the in the uk and you went back to home for will then you will be given the opportunity to apply for a distinguished international associate program and and strengthen in your collaboration with your previous supervisors or previous friends in the uk friends as researchers for sure so uh i think we take in more time in this discussion and thank you shane and thank you all of you for this wonderful discussion and more knowledgeable discussion uh i think massimo is not here yet were put from on hold and we will go to doctor syed ali zaidi from the university of leeds he was [Music] my ex core supervisor in a phd in the university of leeds uh and he is going to talk about uh and the title of his focus sensing the pulse of a city the practical consideration for it deployment and role of machine learning at the edge i think that i thought of this talk is very interesting uh especially for those who are interested in smart cities and when we say smart cities it takes all the parts of engineering units when we talk about ashima university we have eight departments with many programs all of them might be interested in this talk doctor ali if you're already the mic is yours okay so uh thanks sally for inviting me uh today to deliver this talk um thanks shane for supporting this initiative through royal academy of engineering um as you guys have uh mentioned uh you know the that uh this sort of program is really a good segue into uh supporting students who have graduated from university of leeds for example in this case and you know uh trying to establish their independent research in a particular area which has national benefits as well associated with it and it enables us to work closely with uh you know um hash might university in this instance so it's really exciting program and very glad to be part of it uh very uh very thankful to the leadership of ashmite university who in first instance uh you know uh sent uh very excellent students like ali haihajani to leads and then you know that that led on to uh all these uh you know initiatives so really thankful uh for supporting these initiatives uh from the hashemite university um i sort of wear the dual kind of hat because i'm an academic at university of leeds as well but i also have very strong ties with spin-offs in the region so what i will be talking about today would be more on practical side of things so how we translate uh some of the research which is done in the lab into a practical product and what are the considerations which we need to have when we are translating those products into uh you know those technologies into real life product or when we're trying to harness the power of technology to uh enable a certain application what are the things which we need to consider uh in order um to do that and you know um how we should approach such a problem and what is the role of machine learning on the edge in the future to enable those sort of applications so without further delay i will share my screen if i can ali am i host now can i share my screen or uh can you guys see my screen yes okay so uh as uh ali said a long title sensing the pulse of a city uh practical considerations for the deployments in uh role of uh amal at the edge uh what i'll be talking about is a project um which we have uh jointly done with a local startup uh known as away technologies and uh this project is funded by um uk government golf tech challenge catalyst award uh so basically the award the way the award works is that some of the local bodies will pitch different ideas which where they need help and support from technology spin-offs to government and then government will fund those in a few of those initiatives and uh then the startups will bid for a part of money to deliver uh the original challenge uh which was put forward by a public body so in this case the challenge was put forward by leeds and york city council and the challenge was to monitor um their social housing in a in an effective manner to support most vulnerable people and improve the health outcomes for those people living in those properties so as a social housing landlord um elite city council and york city council has a responsibility to maintain uh their properties at a certain standards and then there are lots of regulations around that and how can they harness the power of technology such as internet of things to do that that was the main aim of the project but the principles would generally apply to other areas such as for example we were talking about agriculture uh you know where you have lack of connectivity or limited connectivity in a rural area you would have similar sort of uh principles applied um or we talk about industry 4.0 in you know some in some contacts some of these principles would apply as well stock is mainly uh we'll start with how do we extract so mainly we'll be talking about low power wide area network technologies lp man technologies uh and we'll be discussing how do we harness maximum value out of lp band deployment because there are things which lp man deployment such as lura or you know sick fox or nbiot in 5g space can deliver uh so what what is the maximum uh value we harness from these sort of technologies we'll talk about what are some pre-deployment considerations so when you want to deploy such a technology at a scale uh say you want to monitor 55 000 homes in leeds or you know you want to monitor maybe more than 55 000 homes in jordan uh i would say so you know it's it's a small scale for jordan i think but yeah i mean when we talk about thousands of devices what are the pre-deployment considerations and then we'll be discussing some deployment considerations what are technical considerations for deploying it uh what are our experiences and what are the main challenges and then i'll be discussing uh role of tiny ml which is a tiny machine learning framework in enabling some of the machine learning challenges on the edge okay so uh we have all seen these sort of vertical initiatives which a lot of cities want to do these days so you know you want to monitor air quality you want to have a sustainable mobility especially post coverage given you know i'm not sure whether you guys follow but in uk these days we have a petrol crisis so that uh puts a lot of emphasis on sustainable mobility uh and e-mobility in the future as well um there's a lot of focus on green urban areas um how do we do that how do we sort of optimize the waste management so all these vertical segments of smart city people have been talking about this for for a while now like you know last decade spent looking at different problems in these domains and most of these as uh you know professor khalid also pointed out require interdisciplinary work so when we talk about energy efficiency of the building uh it's not just electrical electronics engineers or wireless coms engineers who can deliver that it would be a close collaboration with civil engineers who understand infrastructure how it behaves what stimulus is uh you know it responds to and how that building fabric changes over time with the lived experience so so there is a lot of interdisciplinary research required in uh enabling all these vertical sectors um what we were interested in uh when we look at all these different verticals as i said we were interested in a particular aspect which was monitoring social housing uh in leeds in new york uh but if we generalize the principle across all these verticals there is a strong belief that data is the new gold basically so you know you want to harness the power of data and use the data to enable all these vertical applications so you know there are lots of local initiatives i won't go through them in very much detail but you know there are projects going on and monitoring for example street lighting making it smart so that it can be repaired in a proactive manner there are projects uh where people are using sensors to monitor flooding potential of uh certain houses and then using uh you know uh the insights from that to improve uh the service delivery but even and if that is required but i mean this belief that data is the new gold led to um led to a sort of a paradigm shift where we have what we have tried what we have started to do is that we have said okay let's collect a lot of data uh forget about what we want to do with the data let's collect all this data uh take it to the cloud and then what we will do uh eventually is that we will analyze data once the analysis is done we will uh sort of get the inference out of it uh whatever principle we want to extract so say let's say house is bad or good we will extract that information in terms of inference and then potentially that will feed into our service delivery whether we do something about that or that house or you know send somebody to inspect that house further for example in this case uh in smart agriculture it would be let's correct a lot of data from crop and then uh you know let's collect data on input and output let's you know mix and match all this data and let's try and maximize the yield without you know having a significant and which inputs do not have significant influence so we can reduce those inputs for example um am i still audible uh early i have some connection issue but uh should be fine okay that's fine thank you so i was saying that you know the the we have started collecting a lot of data without really knowing the value of that data so we collect a lot of data with the hope that there is some value in this data uh now it is very difficult to predict the value of the data uh you know you would say it's a chicken and egg problem we don't know what to collect unless we collect it and analyzed it and have extracted some value from it uh so you know i would call this approach of collecting the data translating into inference and then integrating it into service and upgrading to new service uh to extract the value as a bottom-up principle where it starts with the collection of the data now if we follow this approach this means that we need to deploy the network uh iot network to enable us to collect all the data at very high fidelity at very high resolution spatial temporarily a lot of low power wide area network technologies are not designed for this purpose so you know for example when we talk about lora or sick fox it's few kilobits it's designed to send alarms it's not designed to monitor a property 24 7. so you know um we are talk we are trying to use those technologies in a sort of a manner for which they were not designed and if we use alternatives to those technologies like cellular or you know the other technologies which we we have available they do not give us the ranges we are interested in so they won't cover kilometers we will end up deploying huge amount of infrastructure end up managing that infrastructure there is a management kpac so pax cost associated with that infrastructure as well um the other problem we have is that the more we collect the more we um understand about the process dynamics so let's say if i you know what we put in people's home is the temperature humidity sensors for example if we are collecting uh collecting temperature humidity data from home uh we do understand when somebody was taking shower when somebody was putting kettle on you know all these kind of side information which is encoded in your sensors reading now that violates the privacy of the person where you have put the sensors in first place to monitor damp and mold growth for example in social housing so uh in those areas or say for worker safety you have deployed some sensors but then you can work out where the workers were how efficient they were you know and the sensors were not deployed to monitor their efficiency it was for their safety purposes for example in context of industry 4.0 now uh you know the more data you collect uh the more chances are that you will end up into a territory where privacy would become an issue uh so the principle we should adapt really and what we adapted in this project was collect just enough data hello yeah sorry so i was going to say that the principle we have adopted is that collect just enough data in better service delivery so rather than starting bottom up start top down start with the value where is the value so let me give you an example what i mean let's say we talk about air quality so if we talk about air quality what we are interested in is how is a measure called air quality index air quality index is a function which measures the pollutant concentration over time and prevalence of concentration levels in a certain area so we are not interested in the individual sensor readings of pm10 pm 2.5 nitrous oxide o3 we are interested

in whether aqi index is good satisfactory moderate or very poor so we are interested interested in index where these readings the raw readings have to be averaged over a different duration so let's see here these readings are averaged over 24 hours some readings are averaged over eight hours so rather than sending the raw data all the time if we can somehow use this knowledge that you know whatever service delivery we need to do is based on aqi and then we can live with collecting this data locally and aggregating it and averaging it over a spatial temporal duration which is reasonable enough for service delivery so let's start with the value where is the value value is in collecting aqi value is in let's say how do we want to use that aqi value we can gamify the travel so you know travel on the routes which will minimize let's say air quality impact uh traffic dynamically we want to notify vulnerable people not to travel on certain routes we want to have call for action days you know all these kind of initiatives which we want to deliver as a service are the value for us and these value can be harnessed through just collecting aqi air quality indexes so we do not need to monitor all the processes all the time and transmit that over the network which would uh if we which if we tried to do would uh end up in a very unrealistic requirements for the network okay so what i'm saying is that rather than starting bottom up uh collect all the data we start top down we work out what do we need to improve the service delivery and then we work out uh from that that what are the requirements for network connectivity and what data we need to collect in our case what we wanted to do was improve uh building energy efficiency for example and we had a lower van network deployed in the city uh so the first in order to do that we need to address some of the pre-deployment considerations and those considerations are mainly around scalability of the deployment uh if you're deploying in 55 000 homes then you want to consider uh questions like what do you want to learn from collecting data to influence the change so what what learnings you want to start you know what is the minimum infrastructure required to gather these uh learnings okay how uh what coverage levels do i require how many this will determine how many gateways for example if i'm buying lura when network in a city i would need what sort of technologies i can use to gather that data as well so basically we started i mean you can answer these design questions once you start from top to bottom approach rather than bottom to up approach again there are lots of technology choices in this space so one size does not fit all uh as we know you would have to mix and match uh you have personal area network land van and then lp van lp man can provide you high range uh for your deployment so for example in rural areas you can cover a few kilometers kind of using laura but uh pan has its own role to play in this space as well so for example our solution for monitoring home relies on having sensors which have bluetooth connection uh and then a home hub which uh basically communicates with these sensors over a bluetooth and then transmits the data over laura at the city scale so you would need to combine the different technologies at in different segment you know corners of this diagram but but the key principle in determining what technologies you want to combine is what do you want to uh basically uh collect in order to improve your service delivery okay so laura today i'm not going to talk about all these three boxes i'm mainly for focusing on lp band technologies which is the box green box here in the diagram um laura van i would um presume that you know most of you have come across this it's quite it's a ism unlicensed band 868 megahertz 433 or and 900 recently 2.4 gigahertz as well and uh the principle is that you can send few kilobits at a very high distance so 22 kilometer line of sight and two kilometer non line of sight with uh very uh you know good link budget and the technology is designed to basically enable smart applications which we have discussed before uh the architecture is very uh is typical that you have end devices which communicate with gateway gateways connected uh are connected you know sort of managed through some enterprise network for example things network is one you have other like chip stack which is open source where you can host that network uh yourself in some cloud service like aws for example and manage it yourself uh so laura is a radio frequency technology it uses pet spectrum uh loaded cycle extended coverage extended battery lifetime it's optimized for performing over a long duration um and uh the architecture is pretty much we are all familiar with and devices gateways and cloud so it's again a cloud centric architecture where you collect data send it uh on the cloud so when we started with deployment of lora in leeds the first question was where to place the gateways how many gateways are there and uh one of the things which we could have considered so this is i'm showing you the current gateway deployment elite but what we did was we analyzed we did some rf surveys we analyzed where the gateways could be positioned and in order to work out the gateway location it wasn't just i mean generally the belief is that higher the location is the better the coverage is but higher the location is the coverage will improve but if you place them on only high locations so say for example red locations this is an elevation map shown in the second diagram then potentially it will be away from all the sensors which you want to cover effectively so in this case the sensors uh which is shown in the diagram are lampposts and uh where we want to monitor smart lighting infrastructure so what you need to consider is when you're planning for coverage is where your and users will be so where the nodes will be located in the city and you need to consider can you find out high enough spots where you can deploy very easily uh which are closer to those nodes as well so it's optimization problem i won't go into the details of how to solve that optimization problem but there is a paper which we wrote some time ago which covers different methods of solving this optimization problem especially because you don't want to deploy this network for one application so let's say this these points i'm showing you are a lamppost but you also want to cover houses in the city as well so when you have this sort of optimization problem where you don't know the location of the sensors the priory then uh what you know ali has been working as part of his phd using the knowledge of point processes for example you can exploit that to optimize uh be a pre-deployment uh you know process of the gateways and work out how many gateways are required so that that's you know the process you follow in order to work out the gateway locations as i said unfortunately higher locations will improve the line of sight coverage and receive signal strength at your devices but unfortunately coverage is not equal to receive signal strength so data rate in uh case of lora and several other technologies is a complex function uh particularly for laura it's a complex function of spreading factor because lora uses spread spectrum uh modulation so it you know spreads the signal based on the distance from the base station so if your device is further away from the base station it takes more time to transmit it uh you can transmit only less amount of data because your duty cycle duration is fixed uh if your device is close enough to the base station then you can transmit at a higher data rate so energy and air time uh basically increases as you move further away but data rate decreases as you move further away so um it's it's not easy to work out from just optimizing this paper uh to optimize uh the coverage in in the entire city so what you need to work out is that at a particular spreading factor level let's say you wanted to cover the entire city at spreading factor level seven which will provide you with the rate of uh five five point four seven okay then uh what is what is the minimum number of gateways required at that spreading factor and that might be much more than what i've shown here uh really so you you can cover the city and you will have rsi at a particular spreading factor level but you would need to work out which spreading factor level you do you require and when working out that you also need to work out what are the platform policies so for example if you are using things network uh they will have something called fair access policy in fair access policy as you can see if you are level 12 you can only send one packet in our roughly you know 0.8 packets average per hour which means you can send one packet per hour and that packet can be eight bytes in size now leeds has a network which was not covered at sf7 uh although it's covered uh you know the coverage is uniform in the city but most of the places we get sf11 sf12 so we can really send one packet per hour of eight bytes and that eight bytes packet has to encode the information of the entire house in terms of its thermal performance so that's really interesting challenge again that and we as we know from information theory we can look at how to encode and compress the packets uh to this extent that we can still get some value out of it but uh without having a you know huge payload size uh so uh you know we are limited to what we can send but that that's a good thing because this means that we cannot send everything and we cannot certainly send raw data over this network to improve the service delivery which means that we can preserve the privacy of the users using this network and basically send only limited amount of data uh which is required to do that and work that out using some of the uh the tools which are all in our basket uh the other problem which we want to do is that you know we cannot plan for when we are deploying in real life we cannot plan for the best case scenario it's always better to plan for worst case scenarios so in this case we are planning for sf12 we don't know what spreading factor we will attain um in the real life so let's plan for the worst case scenario and uh basically deploy the system once the system is deployed we have analyzed the data coming out so we can measure rssi levels from this network uh then we can work out that you know if we plot a cumulative distribution function for let's say snr we can work out 90 of the time the snr was above say minus 15 db which means that i can operate at sf10 in this area rather than sf12 for example which i planned for and once i have this information i can uh you know if i have designed my devices in an appropriate manner so i've provisioned firmware upgrades over the year then i can send that firmware upgrade to my device and upgrade it to transmit at sf10 rather than sf12 so we can do all that but the design has to be aware that we plan for worst case but we have provisions in the design to upgrade once we have observed what uh performance it shows in the uh field basically uh again then um there are things you can do to augment the coverage so let's say if you wanted to cover it at a certain level there is a possibility of deploying indoor lora gateways which basically allow you to augment the coverage uh provided by outdoor gateways and you can get typically 680 to 100 1000 meters on on a very simple indoor gateway from uh you know such as this one provided by things network so what we have learned from all this bring algorithms closer to the data so basically run the algorithms on the edge device inside the home for example to come to only get say triggers or you know alarms for example in in our case which gets which you can easily send with very little data payload available to you in a city uh and uh basically that has a lower energy penalty for end devices so you uh you do not consume lot of energy on the device side as well and you can also uh you know ensure privacy and security now in order to do that basically uh you know you need to think about how the current cloud-centric architecture needs to change so currently as i said and as i have presented so far the architecture is you take everything from end device bring it on uh to the cloud you analyze it there uh what i'm saying is that you take everything you analyze it at edge you uh sort of work out summaries or uh you know some sort of a descriptor and then you only share that descriptors with your cloud site and those descriptors can again be used for machine learning in the cloud uh so this brings me on to words ml on the edge uh and as i've mentioned some of these reasons from the previous application uh that you know the reason for doing ml on the edge is one power efficiency um you know most iot devices are battery powered uh they have you know fixed power supply either doesn't exist or power source is not not co-located with where we want to do the monitoring or plugging into mains is costly or you know you cannot move the device uh around as well so it's either power consideration uh there's a latency consideration that you know if you're transmitting everything and then making inference in the cloud and then pushing let's say if you were doing actuation as well as a part of the process if you're pushing that inference back to the end user then the delay encountered in doing that process is not deterministic and you don't have control over it because your information is being sent over internet which introduces random delays so it's better to do this sort of processing on the edge rather than sending it on to the cloud as you can see from uh from very simple results presented here privacy we have touched upon when you have raw data from the sensor it you know the more parts it traverse on internet the more sort of you know chances are that it can be compromised and you can you know lose the trust of your customer uh also there's no transparency so when you're running an algorithm on the edge at least consumer knows that what's going out of his house and you know why that is going out of his house so there's a transparency about what what's being shared and why that is being shared and which applications are authorized to use that shared data if you run these things in the cloud he does he or she doesn't have a transparency you know so he doesn't know that how you would be using his information to make the inferences so there is no explainability there is no uh transparency as such in doing this connectivity again as we have seen in the previous example if you want to deploy a huge amount of infrastructure it's difficult and costly to maintain and make you know such an infrastructure and provide management for it uh it's easy local inferences reduce reliance on the connectivity so let's say if you were operating in a rural area and if your ash device was uh you know good enough to make inferences then you do not need to really have a backhaul connectivity for your edge devices at all you can make uh you know a small uh edge network which can between itself you know some of the gateways coordinator inferences and provide uh reliable inferences uh with the reliable latency to um uh whatever application you're trying to design without relying on connectivity at all so we n

2021-10-07 00:54

Show Video

Other news