Progress 09/01/18 to 08/31/23
Outputs
Target Audience:We focused on a variety of stakeholders during the execution of this grant. Initially, the focus was on the recruiting and training of graduate students for this grant and a parallel grant for the NRI. Wegot engineering and animal science students together virtually and physically to immerse them in each others' work and the prior literature in their respective and shared fields.The students engaged immersedthemselves in literature searches and visiting the farming and engineering labs at their respective universities to better understand the problems and existing solutions of the cattle industry.We added a post-doc to the team and utilized some existing post-docs on related projects to help explore the interdisciplinarity. We had a variety of discussions with the dairy industry,the Ag Extension, and Dairy Management, Inc., to make sure we're solving meaningful long-term problems in returning indiviudalized attention to the large-scale animal agriculture roduction venues, but ultimately decided the work was a bit premature to hold an open workshop. Instead, we plan to have more focused one-on-one discussions wtih early adopters at the end of the NRI grant. With a comprehensive and interdisciplinary literature review from multiple perspectives, targeted the larger research community by developing a set of three separate survey articles covering cyber-animal systems, sensors for biological monitoring of cows, and in-dwelling robotic devices for cow health monitoring. We also expanded out audience to include local future farm community members through anonline interaction with 4-H during covid lockdown. This includes local high schools and middle schools we are including in our outreach efforts to promote STEM and the science of farming. With the survey articles in the review pipeline for the broader research community, we then focused on the more narrow disciplinary audiences in engineering, computer science and animalscience through the individual publishings of the scientists. The graduate students publishedquite regularly in both animal science and engineering/computer science venues, making considerable contributions to IoT for animal ag, modeling for animal nutrition and some methane modeling, as well as security algorithms for authentication-free malware detection. Changes/Problems:We had a change in availability of the test animals at Virginia Tech due to the delays in hiring students at Purdue (caused largely by Coivid-19 and immigration delays) that forced us to re-group and piggyback some experiments on an experiment with beef cattle at Purdue. This let us test our system with the additionof methane detection to examine climate impacts and compare with a priori methane models. Although many of the cannula experimental trials had corrupted data due to the aggressive nature of the beef cattle, we learned some new things about the challenges of the beef industry and the importance of our work on in-dwelling sensing systems (the companion NRIaward). Wealso gathered some tantalizing information on methane concentrations insidethe rumen relative to methane volume production. Although we couldn't establish ground truth on total methane production, we have enough information for a follow-up proposal. What opportunities for training and professional development has the project provided?The leadgrad student at Purdue (in Engineering Technology) graduated with her PhD and joined the Purdue faculty in Agricultural and Biological Engineering. How have the results been disseminated to communities of interest?Threeimportant review papers have been published in both premiere animal science and engineering journals. We have also had about 20 additional discipline-specific scientific papers published across the specturm of six sub-projects: RumenScope, RumenSense, RUMENSoft, RumenComm, RumenNet, RumenStealth. A few more papers are in preparation. What do you plan to do during the next reporting period to accomplish the goals?
Nothing Reported
Impacts
What was accomplished under these goals?
During this final no-cost extension, we built a small number of nodes for testing on cannulated Purdue beef cattle, as opposed to the dairy cattle that were the initial focus of the study. Thereasons for this are three-fold: reliability considerations, industry stakeholder input, and convenience. As we mentioned in the previous annual report, convenience was a key driver because we didn't have the budget to create an entire new study of our own choosing. Instead, we had to piggyback on another study and we chose to useProf. Jon Schoonmaker and a study he was performing with a Latin Square diet trial on beef cattle. Beef cattle were attractive to test with our system because of industry comments on the viability of less-docile beefcatlle with our computerized and sensorized cannula covers (which is likely to be a factor for in-vivo robotic sensors, as well). Dairy cows are more acclimated to humans and are less irritated by changes to their cannula. Also, we have developed a power-scavenging approach to eliminate batterychanges and external recharging and the shorter lifetimes of beef cattle reduces the lifetime of the devices. We gathered data from our indwelling node (minus the RUMENS robot, which is finishing up this year on the NRI grant) using the cannulated beef cows. Wetried to cover cows with different diets, but most of the cows were too aggressive to the modified cannual covers to gather viable data. In post-analysis, we discovered much of the data was corrupted, due toet he cows trying to knockthe covers off, as practioners had warned us. Goal 1: Wedeveloped thedecision layer for the CPS system to incorporate, in a modular way, new models and modling techniques into the monitoring of theherd to provide individual animal estimates of more detailed indicators of health, welfare, and productivity. While our primary goal was development of the infrastructure, rather than individual cow models (because we were limited to cannulated cows), we demonstrated we could gather and upload data to an IoT dashboard with our nodes on individual cows at Virginia Tech. We then demonstrated on individual cows at Purdue, we could flexibly add sensors to the nodes and monitor use-specific metabolic indicators while computing various published low-data-rate models based on averaged animal data. We were not able to generate real-time reformulations of feed supplements other than thosefrom previously-published low-data-rate models. Although this was a secondary sub-goal, we were not able to capture enough variation in cow data over diet changes, due to the aggressive behavior of the beef cows adn the lack of ground truth. Goal 2: Wedemonstrated in two separate experiments the collective ability to flexibly gather and process per-animal information withmultiplesensing modalities, locally pre-process data on active edge nodes, and combine individual data to allow for the reconfigurable analysis of animal and herd metrics of interest. This provided data movement from robotic edge nodes to the cloud via intermediate nodes. Goal 3: Based on commerical-off-the-shelf CPUs and some commercial sensors (plus some novel sensors developed in the lab at low technical readiness levels - TRL), we developed custom collar nodes with in-vivo to ex-vivo communications, custom robotics sensor nodes for in-vivo data gathering, and integrated commercial IoT cloud services. We did not integrate an automated feed delvery device because they have become commercially available and we manually inputindividual cow weights. Goal 4: We linked the physical, network and decision layers in a proof-of-concept demonstration, but we were not able to develop new knowledge on individual cows based on the high-data-rate capabilties of our custom nodes. We showed we could gather and communicate data, but we did not gather a sufficient volume of data over different feeding regimens to input to our new modeling approaches. We showed the new modeling techniquescould work on simulated high-data-rate data, but our high-data-rate real-time data was not sufficiently diverse nor reliable to produce new knowledge from the new modeling techniques. We hope to show this infuture applications of the node and networkartifacts plusmodeling techniques.
Publications
- Type:
Journal Articles
Status:
Published
Year Published:
2023
Citation:
Upinder Kaur, Rammohan Sriramdas, Xiaotian Li, Xin Ma, Arunashish Datta, Barbara Roqueto dos Reis, Shreyas Sen, Kristy Daniels, Robin White, Richard M. Voyles, Shashank Priya,
Indwelling robots for ruminant health monitoring: A review of elements,
Smart Agricultural Technology,
Volume 3,
2023,
100109,
ISSN 2772-3755,
https://doi.org/10.1016/j.atech.2022.100109.
|
Progress 09/01/21 to 08/31/22
Outputs
Target Audience:The graduate students we are working with have been on the project for several years, now, and are well versed in their fields of expertise (classes completed) and in their contributions to the project. The engineering students have all learned a lot about animal science and vice versa. The graduate students are publishing quite regularly in both animal science and engineering/computer science venues. We are making considerable contributions to IoT for animal ag, modeling for animal nutrition and some methane modeling, as well as security algorithms for authentication-free malware detection. The review articles have been submitted and two are accepted. The third is under review. We have been reaching out to the local 4-H organizations for summer courses in origami robots as sensor carriers. Changes/Problems:As noted, we need to augment data collection at Virginia Tech with some new experiments at Purdue. Becasue we don't have the budget to start an entire battery oftests on our own, we are planning to bring Prof. Schoonmaker into the project because he is running a Latin Square feeding test with cannulated cows that we can piggy back onto. He has sutiable IACUC protocols and we plan to run the tests late next fiscal year to verify the data and data models. Wemay need to request a no-cost extension because new cows will have to be cannulated and the gathering and analysis of the data may extend into the following academic year. What opportunities for training and professional development has the project provided?We have strongly focused on writing improvement among all our students and the work has paid off, nicely. They were very productive with conference and journal papers this year, across multiple venues in animal science and engineering. Several students are nearing graduation across all sub-projects. The RumenSense, RumenScope, RumenComm, RumenStealth and RumenNet sub-projects are all producing high-quality graduates with cross-disciplinary skills in animal science, engineering science and computer science. How have the results been disseminated to communities of interest?This year we put a major focus on several publications in both animal science literature and the engineering and computer science literature across the RumenSense, RumenScope, RumenComm, RumenStealth and RumenNet sub-projects. We had 8 new publications and 1 elevated to 'published'. These papers had a strong level of collaboration among the sub-projects and the universities and are reporting good progress. What do you plan to do during the next reporting period to accomplish the goals?We are in the process of developing a new plan to supplementProf. White and Virginia Tech experiments. Since Robin was not able to get a no-cost extension, she is trying to help us as best she can, but she is limited with the scale of tests she can perform that this project can benefit from. We will continue to work with her and Virginia Tech, but need to bring a new professor onboard atPurdue to add some tests at the Purdue Dairy Barn. We are bringing onboard Prof. Jon Schoonmaker to replace Prof. Luiz Brito as he has a major food trial coming up next year with cannulated cows that we can piggyback our testing onto.
Impacts
What was accomplished under these goals?
This year has been very busy with publications and refinement of the core ideas developed last year. We also began integrating professor Luiz Brito into the work this year, as we noted Prof. Donkin has left Purdue for an administrative role. Toward goals 1 and 4, we reported last year that we had begun some preliminary field tests at Virginia Tech to gather richer data on the impact of feed and for the comparison to the popular Molly Model. Profs. Sundaram and Chiu refined the models for novel animal nutrition modeling for the RumenNet system with help from Prof. Brito. In fact, grad student Lei Xin was a finalist for the best paper competition at the American Controls Conference on this work! We got a littlebit of data with another test at Virginia Tech with Prof. Robin White. She was running a small Latin square feeding test with cannulated cows and we sent a team to Blacksburg to gather in vivo data at very high sampling rates, including new data on Rumen methane from inside. The standard sampling rates have been so low that there is not enough data for the new models that Profs. Sundaram and Chiu are developing. The RumenNet system is running orders of magnitude faster than conventional approaches and generating equivalent volumes more data. We got two solid days of data on two cows, but a data recording glitch -- apparently related to confusion on training the VA Tech students to operate the in vivo system -- resulted in a loss of data on the lastthree days. As noted last year, the Molly Model is out of date and no longer supported, so there is an acute need to develop and test these new models with the high-throughput sampling. Unfortunately, Prof. White was not able to submit her no-cost extension for next year, so she has had to focus future testing on other studies that will be lessbeneficial to this project. (More on that in 'changes' section.) Toward goal 2, we continue to gather data and refine the models for the RumenScope and RumenComm sub-projects with Profs. Sen and Brito. Several new papers on the electromagnetic attenuation models were published and we gathered more data on the 'visible rumen' data set with stereo video cameras to analyze rumen motion. We will report more on this in the next report as we continue to analyze the data. Toward goal 3, the RumenStealth sub-project in cybersecurity for the data network has made major refinements of the malware detection paradigm and published several more papers on the general technique for enhancing security for animal agriculture in thepresence of malicious actors. We have been generalizing the technique to robotic edge nodes that allow software updates. Wehave been surprisingly successful at detecting previously unseen examples of rogue malware executables -- to thepoint that we need to refine our writing style as some reviewers have doubted our success rates. Prof Shreyas Sen and his group have designed a new custom communications board for testing of the quasistatic method for low-radiative power body sensor networks. We think this will be highly effective for future medical robotics applications for animals.
Publications
- Type:
Conference Papers and Presentations
Status:
Accepted
Year Published:
2022
Citation:
Upinder Kaur and Richard M. Voyles. 2022. CASPER: Criticality-Aware Self-Powered Wireless in-vivo Sensing Edge for Precision Animal Agriculture. In Proceedings of the 20th ACM Conference on Embedded Networked Sensor Systems (SenSys '22). Association for Computing Machinery, New York, NY, USA, 11231129. https://doi.org/10.1145/3560905.3568419
- Type:
Journal Articles
Status:
Accepted
Year Published:
2022
Citation:
Datta, Arunashish, Kaur, Upinder, Malacco, Victor, Nath, Mayukh, Chatterjee, Baibhab, Donkin, Shawn S., Voyles, Richard M. and Sen, Shreyas, "Sub-GHz In-Body to Out-of-Body Communication Channel Modeling for Ruminant Animals for Smart Animal Agriculture," in IEEE Transactions on Biomedical Engineering, vol. 70, no. 4, pp. 1219-1230, April 2023, doi: 10.1109/TBME.2022.3213262.
- Type:
Journal Articles
Status:
Accepted
Year Published:
2022
Citation:
Upinder Kaur, Rammohan Sriramdas, Xiaotian Li, Xin Ma, Arunashish Datta, Barbara Roqueto dos Reis, Shreyas Sen, Kristy Daniels, Robin White, Richard M. Voyles, Shashank Priya,
Indwelling robots for ruminant health monitoring: A review of elements,
Smart Agricultural Technology,
Volume 3,
2023,
100109,
ISSN 2772-3755,
https://doi.org/10.1016/j.atech.2022.100109.
- Type:
Journal Articles
Status:
Published
Year Published:
2022
Citation:
Chan Su Han, Upinder Kaur, Huiwen Bai, Barbara Roqueto dos Reis, Robin White, Robert A. Nawrocki, Richard M. Voyles, Min Gyu Kang, Shashank Priya,
Invited review: Sensor technologies for real-time monitoring of the rumen environment,
Journal of Dairy Science,
Volume 105, Issue 8,
2022,
Pages 6379-6404,
ISSN 0022-0302,
https://doi.org/10.3168/jds.2021-20576.
- Type:
Conference Papers and Presentations
Status:
Published
Year Published:
2022
Citation:
Kaur, U., Ma, X., Voyles, R.M., Min, BC. (2022). Malware Detection Using Pseudo Semi-Supervised Learning. In: El Yacoubi, M., Granger, E., Yuen, P.C., Pal, U., Vincent, N. (eds) Pattern Recognition and Artificial Intelligence. ICPRAI 2022. Lecture Notes in Computer Science, vol 13364. Springer, Cham. https://doi.org/10.1007/978-3-031-09282-4_31.
- Type:
Conference Papers and Presentations
Status:
Published
Year Published:
2022
Citation:
U. Kaur, Z. B. Celik and R. M. Voyles, "Robust and Energy Efficient Malware Detection for Robotic Cyber-Physical Systems," 2022 ACM/IEEE 13th International Conference on Cyber-Physical Systems (ICCPS), Milano, Italy, 2022, pp. 314-315, doi: 10.1109/ICCPS54341.2022.00048.
- Type:
Conference Papers and Presentations
Status:
Published
Year Published:
2022
Citation:
L. Xin, G. Chiu and S. Sundaram, "Finite Sample Guarantees for Distributed Online Parameter Estimation with Communication Costs," 2022 IEEE 61st Conference on Decision and Control (CDC), Cancun, Mexico, 2022, pp. 5980-5985, doi: 10.1109/CDC51059.2022.9992593.
- Type:
Conference Papers and Presentations
Status:
Published
Year Published:
2022
Citation:
L. Xin, G. Chiu and S. Sundaram, "Learning the Dynamics of Autonomous Linear Systems From Multiple Trajectories," 2022 American Control Conference (ACC), Atlanta, GA, USA, 2022, pp. 3955-3960, doi: 10.23919/ACC53348.2022.9867533.
- Type:
Conference Papers and Presentations
Status:
Published
Year Published:
2022
Citation:
L. Xin, L. Ye, G. Chiu and S. Sundaram, "Identifying the Dynamics of a System by Leveraging Data from Similar Systems," 2022 American Control Conference (ACC), Atlanta, GA, USA, 2022, pp. 818-824, doi: 10.23919/ACC53348.2022.9867413.
|
Progress 09/01/20 to 08/31/21
Outputs
Target Audience:The primary target audience of this report period hasbeen the graduate students that are progressing on their research topics. Unfortunately, ourAnimal Science post-doc will be leaving us as his time is up and he is moving on. This was predicated by his advisor, Prof. Shawn Donkin, accepting a job at another university, so we are losing two key animal science collaborators, here at Purdue. (But we have identified replacements, already.) To continue to broaden our focus to the largerresearch community, we have completed the first drafts of three separate survey articles covering cyber-animal systems, sensors for biological monitoring of cows, and in-dwelling robotic devices for cow health monitoring. (Details of the papers will be reported separately.) Finally, a tertiary audience is the local farm community and local 4-H. This includeslocal high schools and middle schools we are including in our outreach efforts to promote STEM and thescience of farming. Changes/Problems:Theaddition of Prof. Shreyas Sen has been major boost to the RumenComm sub-project. Unfortunately, we are sad to report that Prof. and Assoc. DeanShawn Donkin has left Purdue for an administrative position at Oregon State University. Since he will not have time to work on research in this new position, we are happy to report that Asst. Prof. Luiz Brito, an expert in observational genomics of dairy cows, will be taking over for Prof. Donkin. We expect equally great things from Prof. Brito! What opportunities for training and professional development has the project provided?Technical writing has been the primary theme for training and professional development for the students and post-docs. Developing the three survey papers for sensors, robots and systems has been a galvanizing effort that has really brought the team together. We also developed a novel approach to mentoring for theCyber-Animal Systems class we plan to deploy this fall with Profs. Marisa Erasmus and Luiz Brito. We have a novel approach to build our students into "Lifelong Teachers" as well as "Lifelong Learners". We think this will be atransformational aspect of training for STEM scientists to go into the world disseminating science to the general public at evey opportunity. It is important for scientists to evangelize the benefits of science and logic at all levels. How have the results been disseminated to communities of interest?We are ramping up our conference papers and journal papers aimed at both animalscience and engineering audiences and we expect the survey articles will be well-received as landmarks in theemerging field of precision animal agriculture (precision livestock farming). One has been published already, the second is in revision and the third will shortly be submitted. We also developed a novel hybrid program on paper-based origami robots for animal agriculture for local 4-H groups at the middle and high school levels. Due to Covid-19, we had to deliver this program remotely, but it was a hit with kids. We teamed with our Agirculture Extension to recruit 4-H clubs from all around the state of Indiana to elect to join. We had several online activities the students could elect to particiapte in and the "Origami Ag Robots" option was one of the most popular. We offered two sessions in which we mailed custom-built origami PaperBot kits out to kids before the online learning session and had them join us online to show them how to build it and the STEM principles of how it worked. In the end, they have a small, folded paper robot (minus the motor), with Purdue logo, that they can keep! We have been quite active in joint publications and have had many joint papers published or submitted on all aspects of our comprehensive work, including RUMENSoft, RUMENComm, RUMENnet, RUMENstealth, and RUMENscope. What do you plan to do during the next reporting period to accomplish the goals?Only one of the suvery journal articles has been accepted to-date, but we have full drafts of the next two, one targeted to the Journal of Animal Systems and the other to a special issue of Smart Agriculture Technology. As mentioned above, we have a multi-pronged plan to deploy virtual sensors based on mathematical models from theliterature and from cooperation between our teams at Virgina Tech and Purdue. Wewill first deploy networks in the research farm environment with: - pre-recorded data - then with literature-derived models of methane and pH - then with custom-derived models from the Purdue modeling team based on Virginia Tech data - finally live sensor data from Penn State and Purdue sensors integrated to the edge and collar nodes We have addtional work to do processing the RumenScope electromagnetic and physical models and expand the database of cow data. Another custom hardware board will be developed to explore a new quasistatic signalling approach for the RumenComm body sensor network for ultra-low power secure communication. Wewil continue the RumenStealth work on malware detection and incororate it from thetest networks into the animal agriculture prototype. Finally, we have developed a novel form of in-rumen locomotion we call "buoyancy gaits." We are working on a novel soft robot tocomplement the work at Penn State on in-dwelling robotic sensor nodes. This builds on the successful deployment of a hybrid modular robot composed of Purdue modules and Penn State modules.
Impacts
What was accomplished under these goals?
Toward goal 1, working with Profs. Shreyas Sundaram, George Chiu, and Shawn Donkin, we have explored novel animal nutrition modeling approaches from the controls community for the RumenNet system analytics. Coordinating with Prof. Robin White, Virginia Tech, we have tested some of her data from animal feed trials to build individualized models of per-cow productivity andwellness and compared them to the existing Molly model for cow nutrition. (Two papers submitted to ACC on this topic.) We have found the existing Molly model is out of date and no longer supported, so there appears to be a real need for new, comprehensive, open-source models. Based on existing literature, we implemented existing models for methane production and rumen pH in our network testbed. We also developed a novel approach to authentication-free cyber security for animal agriculture CPS networks. We developed a semi-supervised training approach to malware detection that has high likelihood of detecting malware injected into nodes in our network architecture. Just likethe STUXnet worm caused damage and havoc to the Iranian nuclear centrifuges a decade ago, we fear rogue malware may target the food supply of future CPS-powered farms using our reference architecture (or other CPS frameworks). Therefore, practical security must be built-in that does not impede the farmers' operations. This enhancement to the RumenStealth sub-project has been published in theIEEE Robotic Computing conference and a more advanced version is under review for CPS week in collaboration with Prof. Min. Toward goal 2, Working with Prof. Shreyas Sen and his team, we have made major strides for the RUMENscope and RUMENcomm sub-projects. RumenScope is the physical and electromagnetic mapping of the cow rumen. We combinedcustom hardware with off-the-shelf cameras and software-defined radios to build internal and external physicalmodels of the cow and cow rumen, merged with ultrasound measurements of tissue layers and thicknesses. Correlated with these physical models, we ran measurements of RF attenuation with a radio transmitter placed inside the empty rumen of a cannulated cow and a receiver scanned around the entire outside of the cow in an ellipsoidal path. From this, we build a virtual electromagnetic model of the cow, compared it to actual attenuation data, and determined the optimal RF signal band and locations for in-body to out-of-body communication. (One paper published in EMBC.) In addition, we used stereo vision to map portions of the inside of the cow rumen during several contraction cycles. Weare currently processing the internal 3-D visualizations of rumen contractions. Toward goal 3, we developed a preliminary set of 5 nodes of the RumenNet CPS reference architecture and achieved end-to-end communication from a single sensing edge node, to3 collar nodes, and up to the cloud node with Prof.Min. The majority of nodes in this test reported pre-acquired cow data from Prof. Robin White's animal nutrition studies at Virginia Tech. (Which she can access using our online dashboard.) Only the edge node was reporting live data for this end-to-end test with Prof Nawrocki, who is developing new sensors for the network. We developed a completely new approach to locomotion and sensor positioning for inside the rumen based on "buoyancy gaits" and a new modular robot that is complementary to the Penn State robot of Prof. Priya. (One paper published in ReMAR.) Toward goal 4, we collaboratively developed a plan for the coming year to gradually merge the models and data from Virginia Tech, the sensors, robots and power scavenging from Penn State, and the sensors, networks, and data models from Purdue into a prototype-scale system that we hope we can use to work with the Dairy Management Institute (DMI) on 4-H education on the future of farming.
Publications
- Type:
Book Chapters
Status:
Published
Year Published:
2021
Citation:
Upinder Kaur, Richard M. Voyles and Shawn Donkin, "Future of Animal Welfare - Technological Innovations for Individualized Animal Care," in Improving Animal Welfare, 3rd Edition, Edited by Temple Grandin, CABI
- Type:
Conference Papers and Presentations
Status:
Published
Year Published:
2021
Citation:
Upinder Kaur, Haozhe Zhou, Xiaxin Shen, Byung-Cheol Min, and Richard M. Voyles, "RoboMal: Malware Detection for Robot Network Systems," in Proc. of the IEEE Intl. Conf. on Robotic Computing, 2021.
- Type:
Journal Articles
Status:
Accepted
Year Published:
2021
Citation:
Chan Su Han, Upinder Kaur, Huiwen Bai, Barbara Roqueto dos Reis, Robin White, Robert A Nawrocki, Richard M. Voyles, Min Gyu Kang, and Shashank Priya, "Invited Review: Sensor technologies for real-time monitoring of the rumen environment," in Journal of Dairy Science
- Type:
Conference Papers and Presentations
Status:
Published
Year Published:
2021
Citation:
Arunashish Datta, Upinder Kaur, Victor Marco Rocha Malacco, Mayukh Nath, Baibhab
Chatterjee, Shawn Donkin, Richard Voyles, Shreyas Sen, "In-body to Out-of-body Communication Channel Modeling for Ruminant Animals for Smart Animal Agriculture," in IEEE Intl. Conf. on Engineering in Medicine & Biology
- Type:
Conference Papers and Presentations
Status:
Under Review
Year Published:
2022
Citation:
L. Xin, L. Ye, G. Chiu, and S. Sundaram. Identifying the dynamics of a system by leveraging data from similar systems. In American Control Conference, 2022
- Type:
Conference Papers and Presentations
Status:
Under Review
Year Published:
2022
Citation:
L. Xin, G. Chiu, and S. Sundaram. Learning the dynamics of autonomous linear systems from multiple trajectories. In American Control Conference, 2022
- Type:
Conference Papers and Presentations
Status:
Published
Year Published:
2021
Citation:
Yuanmeng Huang, Jonathan Miller, Upinder Kaur, Yubing Han, Ram M.S.Ramdas, Shashank Priya, and Richard M. Voyles, `Hybridization Through Modularity: Exploring Complex Modes of Locomotion with a Bag of Robotic Modules ,' in 5th IEEE/IFToMM Intl Conf on Reconfigurable Mechanisms and Robots, 2021
|
Progress 09/01/19 to 08/31/20
Outputs
Target Audience:The main target audience has, again, been the students as we are now up to speed, from a research perspective. The new grad students are excellent and the Animal Science post-doc is also quite good. We have started to broaden our focus to the more general research community and, to a lesser extent, the local farm community, including loical farmers and local 4-H. Wealso reached out to high schools and middle schools to start to promote the work we are doing. Changes/Problems:As we noted previously, we brought onboard a new expert in body sensor networks. This has resulted in a significant boost to the project and the collaboration is working out extrememly well. What opportunities for training and professional development has the project provided?The students and post-docs are developing in their graduate careers and have begun to develop papers and target them to specific conferences. As the students are all relatively new, they are in need of develping their technical writing style. How have the results been disseminated to communities of interest?Several papers have been written for both agriculkture and engineering conferences and some journal articles. We also have completed the chapter on the "Future of Animal Welfare" from a technology perspective. Finally, we have visited local midele and high schools and discussed sensors and robotics for animal agriculture. What do you plan to do during the next reporting period to accomplish the goals?We have decided to publish a series of review articles in the areas of the grant. Namely, we are targeting a 3 or 4-paper suite of survey articles for the Journal ofDairy Science to cover sensors, in-dwelling robots and active devices, and CPS systems, plus a similar systems survey paper in an engineering journal. We are also going to expand our electromagnetic testing protocols for in-cow testing and the development of a database of cow data for RUMENscope. This will include stereo-pair imagery to extract intenal 3-D shape of the rumen. We must expand the definition of the RUMENstealth cybersecurity approach we are developing for malware detection. Although semi-supervised methods are less dependent on trianing data -- a major plusfor this effort -- there is still a high dependence on training. We must reduce this dependence.
Impacts
What was accomplished under these goals?
We made major advancements in the networking infrastructure and the networking protocols for security and integrity, relavant to the agriculture application. Weexpanded the team to include an established expert in the field of body sensors networks. This expert is a new junior faculty member that works in human medical applications of body sensor networks and is eager to explore the realm of animal body sensor networks. We have been laying the ground work and simulations for a new form of quasistatic galvanic network that will be ultra-safe for the animals. A common concern around body sensopr networks is the potential danger of RF radiation near sensitive parts of the body. Not much work has been done to study the electromagnetic properties of the cow and we are building the artifacts we need to create "RUMENscope" a new observatory for the inside of the cow. We have run some initial experiments in the measurement of attenuation properties of cow tissues by placing a software-defined radio transmitter inside the cow and gathered signal strength measurements all around the cow. We also are attempting to gather the first-ever 3-D video of the cow rumen undedrgoing contractions. The "RUMENstealth" project is a new, authentication-free, cybersecurity approachto discover malware on the active sensors and robots inside the cow, before the malware has a chance to execute. It is not practical for farmers to "login" to each and every cow to authenticate for security purposes. Instead, it is important to develop trustworthy methods to maintain the integrity of hte animal network to protect the animals. This approach uses semi-supervised learning to determine "good" or "bad" software updates prior to execution. This will be an important component of the CPS Reference Architecture for Animla Agriculture. We continue to develop the histamine, pH and other chemical sensors necessary to gather data from the cows to develop the models of nutrition, health and welness for individual animals. We have been comparing the industry-standard "Molly" modelfor dairy cows' milk quality to new estimations of individual Kalman Filter based models.
Publications
- Type:
Book Chapters
Status:
Awaiting Publication
Year Published:
2021
Citation:
Improving Animal Welfare, 3rd Edition, CABI
- Type:
Journal Articles
Status:
Under Review
Year Published:
2020
Citation:
Impedimetric, PEDOT:PSS-Based Organic Electrochemical Sensor for Detection of Histamine for Precision Animal Agriculture, IEEE Sensors Letters
- Type:
Conference Papers and Presentations
Status:
Published
Year Published:
2020
Citation:
Organic Electrochemical, {PEDOT}:{PSS}-Based Impedimetric Histamine Sensor, ECS Meeting Abstracts
- Type:
Conference Papers and Presentations
Status:
Under Review
Year Published:
2020
Citation:
Malware Detection Using Pseudo Semi-Supervised Learning, IAAI Conference
|
Progress 09/01/18 to 08/31/19
Outputs
Target Audience:The target audience, tothis point, has primarily been the students we have been attempting to hire for the project. The students engaged have been immersing themselves in literature searches and visiting the farming labs at their respective universities to better understand the problems and existing solutions of the cattle industry. We have had mulitple setbacks in hiring of graduate students, including immigration delays, a student that got rejeted at O'Hare Airport, and even internal processing delays, but we are making up for the hiring problems by leveraging existing ongoing and peripheral projects to keep the progress mobing forward. We have a new post-doc, some existing post-docs on related projecdts, and are bringing on a new faculty member to help with the work load. Now that we are gettingpast these start-up problems, we intend to re-focus our attention on the dairy industry and the Ag Extension, including a partnership with Dairy Management, Inc., to begin interacting with actual farmers to make sure we're solving meaningful long-term problems in returning indiviudalized attention to the large-scale animal agriculture roduction venues. Changes/Problems:Our department and college are in a growth phase, which is causing a shortage of graduate students, increasing the sensitivity to individual student decisions and fluctuations. Our acceptance policies are lagging behind the growth in propsoal activity and student hiring and post-docs, which we normally rely on to bridge difficulties,are harder to come by at this time, as well. We are excited to bring Shreyas Sen onboard, a new junior faculty member with valuable expertise in body sensor networks. We are also trying to bring onboard another post-doc. What opportunities for training and professional development has the project provided?
Nothing Reported
How have the results been disseminated to communities of interest?As noted, we began preparing papers for submission to venues in 2020. These include the ICCPS (CPSweek) conference and a book chapter for Temple Grandin's revised 3rd edition of the book "Improving Animal Welfare". What do you plan to do during the next reporting period to accomplish the goals?The new graduate student we hired is critical to the project as she will act as the point person for the integration of the various sub-parts. As mentioned above, we are finalizing an agreement with a junior faculty member to join th project to provide substantial expertise on body sensor networks for in-vivo sensing and the safe extraction of data from live animals. This will begin with an RF-based aproach, but we are workin out a schedule that will migrate to low-frequency methods of signal transduction. As this is established, we will create a sub-team across the universities to examine the computational substrate facilities necessary to host the various forms of communicaitons and the analytics necessary to suport them. We will define the interfaces and capabilties of the various platforms for sensing, power, networking, and actuation. We plan to flesh out the refernce architecture with the primary focus of uniqueness at theedge layer (in-vivo animal sensors). The edge layer in animal agriculture we believe requires special constructs and constraints, so it will be natural and successful to define the "analytics" to support and conform to these affordances. We have also started to analyze the data gathered by Robin White's group at Virginia Tech using the information-theroetic approaches for modle-based data analysis of Shreyas Sundaram and George Chiu. We are examining the structure of the data and applying comparative metrics to model extraction by variousstandard means such as Kalman Filters and Model Predicitive Control. We expect to have preliminary recommendations by next reporting period and a plan for testing and verification.
Impacts
What was accomplished under these goals?
As mentioned previously, we had great difficulty getting students onboard for a variety of reasons last year, and that has extended, partially, to the second year, with a second student that committed to coming to Purdue, but decided to stay put at the last minute. Nonetheless, we did bring onboard a new very talented female engineering technology student at the beginning of the summer that is picking up the slack. We also have identified a new facutly member that can add some "instant" expertise and students in the area of body sensor networks that will accelerate our progress in the networking area. We are negotiating a statement of work and expect to have a finalized agreement in place early in the next reporting cycle. The network architecture of the proposed system involves three layers which include the farm level (cloud storage layer for the entire farm), the sub-herd level (collar nodes that move around with the animals), and the individual animal level. Upon analyzing our network needs and based on the literature review of alternatives, we decided to leverage existing work on a crop agriculture project several of the co-PIs are involved in to establish the cloud and collar layers. Because 5G is not widely available even in urban environments, let alone rural environments, we plan to use zigbee as the primary protocol for layers 1 and 2. Zigbee is preferred because we will need peer-to-peer networking between different collar nodes in order to track such things as individual animal activity relative to sub-herd average activity and to track which animals belong to various sub-herds that might be in close proximity. We're borrowing the arduino/zigbee architecture develped by Min, Voyles and their post-doc in a crop agriculture project to prototype layers 1 and 2 of the network. We used an Arduino with Xbee comm module to gather and upload data to the ThingsBoard open-source cloud platform. Without the in-vivo layer 3 (whichshould be ready in about 15 months or less), the characteristics of the network architecture are not too different from other projects we are developing, so they make a reasonable surrogate.
Publications
- Type:
Conference Papers and Presentations
Status:
Other
Year Published:
2020
Citation:
Developing submission for 2020 Intl Conf on Cyber-Physical Systems
- Type:
Book Chapters
Status:
Other
Year Published:
2020
Citation:
developing new chapter for Improving Animal Welfare: A Practical Approach, 3rd Edition
|
|