Difference between revisions of "Projectlist"
Line 1: Line 1:
 
=Part II/ACS projects=
 
=Part II/ACS projects=
−
* ''Federated blockchain'': The goal of this project is to design and implement a [https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3236753 federated blockchain], where gateways are used to exchange data between blockchains to overcome differences in jurisdictional requirements. Specifically, each jurisdiction (country or a region of a country) has its own requirements for disclosure of project details for carbon credits. This makes it difficult to define a single blockchain transaction type that can be used globally. With this approach, a chain can store sector- and location-specific details of carbon sequestration from nature-based solutions. The gateways would provide translation of attributes between chains. The solution should be general, scaleable to global scale, and deployable using current mainnets such as for Tezos and Algorand.
 
  
−
* ''Trusted image capture'': The goal of this project is to link image capture from trusted hardware devices, such as [https://ieeexplore.ieee.org/document/9302967 Azure Sphere], to a global file store, such as IPFS, with summaries posted to a blockchain. This would allow us to trace an image to its creator with an unbroken chain of trust. Students who have some background working with microcontroller-based single-board devices, such as a Raspberry Pi, would be preferred. An alternative would be to develop the solution using an integrated hardware/software platform, such as iOS, and use iOS APIs to prove that the captured image, its time stamp, and device orientation had not been tampered with.  
+
* ''Reinforcement Learning for Bi-directional EV charging'': Today's EVs are mostly one-way, that is, they charge, but they cannot supply energy back to the grid. However, [ https://www.canarymedia.com/articles/ev-charging/is-vehicle-to-everything-charging-ready-for-prime-time vehicle-to-grid charging] is far more rewarding for home owners, especially those with their own solar panels. But when exactly should the EV be charged and when should it be discharged? This is a complex problem that is determined by when the EV is present at home, the next day's travel plans, grid requirements, and so on. The goal of this project is to use reinforcement learning to come up with EV charge/discharge control, similar to the approach proposed [https://ieeexplore.ieee.org/abstract/document/9345625 here], but extending the use cases to make it more realistic.  
  
−
* ''Speeding up forest simulation'': The [https://iopscience.iop.org/article/10.1088/1748-9326/aaaacc state of the art] in forest simulation is an agent-based model, such as [https://github.com/TROLL-code/TROLL  TROLL]. Here, a software agent simulates each tree in a forest. This process is highly parallelisable, and the goal of this project is to exploit the inherent parallelism to greatly speed up agent-based simulations. (This project is now taken)
+
* ''Trunk diameter detection for complex trunks'': In recent [https://svr-sk818-web.cl.cam.ac.uk/keshav/wiki/images/0/01/Mobisys21postersdemos-final85_copy.pdf work] we have used mobile phones with LIDAR to measure trunk diameters. However, our algorithm only works for nearly cylindrical trunks. The goal here would be to build on this work to tackle more complex trunks, such as those with burls, lianas, or multiple trunks.  
  
−
* ''Durable storage of primary observational data'': Observational data from trusted sources, such as satellites and trusted cameras (see above) are voluminous (terabytes to petabytes) but need to be stored for decades. This project explores storage and indexing of such data. Ensuring immutability of the data through a blockchain link would also be desirable. (This project is now taken)
+
*''A solar PV layout and sizing tool'': Residential solar PV is rapidly being deployed on rooftops around the world. In prior [https://svr-sk818-web.cl.cam.ac.uk/keshav/wiki/images/8/8b/Sizing_multiple_roofs-5_copy.pdf work] we have studied the sizing of solar and storage (i.e. how many panels and how much storage to purchase) to meet a certain level of grid independence. This work, however, does not take into account the realities of rooftop PV, which includes limitations on where panels can be placed (avoiding skylights, for example), shadowing from rooftop protuberances and nearby trees, and self-shadowing from adjacent panels. The goal of this project is to use rooftop images, either from a satellite or a drone, to create a 3D model of the rooftop, then do an [https://www.researchgate.net/profile/Rawad-El-Kontar/publication/344453369_Optimal_Efficiency_and_Operational_Cost_Savings_A_Framework_for_Automated_Rooftop_PV_Placement/links/5f7740cc299bf1b53e09526e/Optimal-Efficiency-and-Operational-Cost-Savings-A-Framework-for-Automated-Rooftop-PV-Placement.pdf optimal placement], subject to a sizing requirement.
  
−
* ''Sensor design'': Many sensors in use today, especially in agricultural contexts, are not suitable for long-term mass-scale forest deployment. Forest environments are less controlled and consistent than agricultural ones, and access to deployed sensors is comparatively limited. Thus, the sensors need to be self-managing, robust and long lived. Sensors may make different trade-offs between energy usage and lifespan. For example, active medium-lifespan sensors (e.g. spectroscopy for measuring soil quality) used during sapling establishment could be deployed alongside passive long-lived sensors designed to last for the expected lifetime of the trees (e.g. temperature and humidity via chipless RFID).Additional sensors may also be considered outside of those normally associated with agriculture, for example to account for dead organic matter, estimate carbon content of soil and litter, and to monitor for poaching, theft, vandalism and possible fraud. Thus, there is ample scope for research into appropriate sensor system design.
+
* ''Trusted image capture'': The goal of this project is to link image capture from trusted hardware devices, such as [https://ieeexplore.ieee.org/document/9302967 Azure Sphere], to a global file store, such as IPFS, with summaries posted to a blockchain. This would allow us to trace an image to its creator with an unbroken chain of trust. Students who have some background working with microcontroller-based single-board devices, such as a Raspberry Pi, would be preferred. An alternative would be to develop the solution using an integrated hardware/software platform, such as iOS, and use iOS APIs to prove that the captured image, its time stamp, and device orientation had not been tampered with.
−
 
 
−
* ''Geolocation in a rainforest'': The goal here is to design a geolocation service in a rainforest, where GPS may be either unavailable or imprecise and the environmental conditions are harsh.  Under such circumstance, the typical 10-30m accuracy achieved in the field would not be sufficient for fieldwork. The goal here would be to implement one of the many systems described for indoor wayfinding, such as using BLE and RSSI, e.g. https://ieeexplore.ieee.org/abstract/document/7943542 , in an outdoor context.  (This project is now taken)
 
−
 
 
−
* ''Drone-based forest biomass and soil analysis'': The goal of this project is to use drone along with a camera and a hyperspectral sensor to create pointclouds of the forest, and to run PLSR or RandomForest to work out which particular wavebands have reflectances that predict soils properties well.
 
−
 
 
−
* ''Linking field data and remote sensing'': How can we integrate diverse spatial and temporal field/point measurements of biodiversity and ecosystem functioning using remote sensing data? The idea is to use ground based sensors to establish ground truth for remote sensing imagery and remote sensing to tie together ground based sensors.
 

Revision as of 17:27, 27 April 2022

Part II/ACS projects

  • Reinforcement Learning for Bi-directional EV charging: Today's EVs are mostly one-way, that is, they charge, but they cannot supply energy back to the grid. However, [ https://www.canarymedia.com/articles/ev-charging/is-vehicle-to-everything-charging-ready-for-prime-time vehicle-to-grid charging] is far more rewarding for home owners, especially those with their own solar panels. But when exactly should the EV be charged and when should it be discharged? This is a complex problem that is determined by when the EV is present at home, the next day's travel plans, grid requirements, and so on. The goal of this project is to use reinforcement learning to come up with EV charge/discharge control, similar to the approach proposed here, but extending the use cases to make it more realistic.
  • Trunk diameter detection for complex trunks: In recent work we have used mobile phones with LIDAR to measure trunk diameters. However, our algorithm only works for nearly cylindrical trunks. The goal here would be to build on this work to tackle more complex trunks, such as those with burls, lianas, or multiple trunks.
  • A solar PV layout and sizing tool: Residential solar PV is rapidly being deployed on rooftops around the world. In prior work we have studied the sizing of solar and storage (i.e. how many panels and how much storage to purchase) to meet a certain level of grid independence. This work, however, does not take into account the realities of rooftop PV, which includes limitations on where panels can be placed (avoiding skylights, for example), shadowing from rooftop protuberances and nearby trees, and self-shadowing from adjacent panels. The goal of this project is to use rooftop images, either from a satellite or a drone, to create a 3D model of the rooftop, then do an optimal placement, subject to a sizing requirement.
  • Trusted image capture: The goal of this project is to link image capture from trusted hardware devices, such as Azure Sphere, to a global file store, such as IPFS, with summaries posted to a blockchain. This would allow us to trace an image to its creator with an unbroken chain of trust. Students who have some background working with microcontroller-based single-board devices, such as a Raspberry Pi, would be preferred. An alternative would be to develop the solution using an integrated hardware/software platform, such as iOS, and use iOS APIs to prove that the captured image, its time stamp, and device orientation had not been tampered with.
Retrieved from "https://svr-sk818-web.cl.cam.ac.uk/keshav/wiki/index.php?title=Projectlist&oldid=1085"