Smart City
(Panotti's Ears) Real time Voice Analytics Co-processor
AP007 »
Speech recognition models have been used extensively on various platforms to provide ease of use, digital smart assistance, and hands-free control. At the cutting edge of this technology, the use of hidden Markov models is common. To improve the computational efficiency of hidden Markov model-based speech recognition systems, various techniques are used, amongst which the Viterbi beam search algorithm is one of the best. However, for large vocabulary speech recognition models with larger beam widths, the beam search algorithm’s sparse matrix operations create a highly constrictive bottleneck. Traditionally GPUs have been used to accelerate such models but with the algorithm's not so parallel nature, GPUs don’t provide an efficient solution and power constraints of IOT devices completely rule them out for Edge level.In this project, we research and formulate an FPGA based Co-processor (RTL level abstraction) to accelerate sparse matrix operation of the beam search algorithm so it can be used on edge devices to revolutionize how we interact with IOT edge level devices.
Industrial
FPGA based Wireless Sensor Networks
AP009 »
Wireless Sensor Networks (WSN) implementation on FPGA connecting multiple nodes with different sensors from analog devices and actuators to make industrial work more independent of human control physically. The data management will be made possible using Microsoft Azure cloud services, making the data accessible as and when required to the controller and the specifications can be tuned at admin end. This project is to make cloud based applications on FPGA and to explore new emerging technologies to make them system independent using FPGA.
Data Management
Cactus Fish
AS007 »
Utilizing Microsoft Azure to implement IoT solutions on DE10
Smart City
Model-Predictive-Traffic-Mangement
AP010 »
We are proposing to build a hardware platform that's able to identify individual vehicles in real-time. The device will be connected to a real-time traffic model that's run on the cloud. Then by placing this device at a strategic location in a city, over time, we would be able to build a very accurate predictive traffic model.
Overall, we expect our project (hardware platform together with the back-end cloud-based software) will provide much better tools for city authorities to manage traffic than what's available today.
Transportation
AI based data-driven approach and hardware accelerators (FPGA) to predict the SoC, SoH and RuL of the LiBs
AP012 »
The advancement in digitalization and availability of reliable sources of information that provide credible data, Artificial Intelligence (AI) has emerged to solve complex computational real-life problems which was challenging earlier. However, Artificial Neural Networks(ANN) need rigorous main processors and high memory bandwidth, and hence cannot provide expected levels of performance. As a result, hardware accelerators such as Graphic Processing Units (GPUs), Field Programmable Gate Arrays (FPGAs), and Application Specific Integrated Circuits (ASICs) have been used for improving overall performance of AI based applications. FPGAs are widely used for AI implementation as FPGAs have features like high-speed acceleration, low power consumption which cannot be done using central processors and GPUs. In Electric-powered vehicles (E-Mobility), Battery Management Systems (BMS) perform different operations for better use of energy stored in lithium-ion batteries (LiBs). The LiBs are a non-linear electrochemical system which is very complex and time-variant in nature. Because of this nature, estimation of States like State of Health (SoH) and Remaining Useful Life (RuL) is very difficult. The goal is to develop an advanced AI based BMS that can precisely indicate the LiBs states which will be useful in E-Mobility. This gives useful information for the prediction of when the battery should be removed or replaced and helps to optimize the battery performance and extend battery lifespan.
Smart City
Project Watchdog
AS008 »
Project Watchdog is an FPGA-based smart home security camera. Existing solutions such as Google Nest and Amazon Ring require an internet connection and a monthly subscription to operate properly. They send the video to an external server and then perform all processing on that server, increasing bandwidth and latency. Furthermore, these devices are rendered useless without an internet connection. Watchdog will perform all video capture and analysis right on the device, regardless of an internet connection. The device will run inference using an AI model that has been trained to identify people and animals. When any people or animals are detected, that snippet will be uploaded to Azure cloud storage, for easy online access to these clips. The video footage will be stored on a micro-SD card on the FPGA board, which can be accessed if the full video needs to be viewed. Ultimately, Watchdog will perform identically to the state-of-the-art consumer solutions, but will use significantly less bandwidth and latency while not requiring an internet connection. The FPGA solution will be paired with a temperature sensor to provide a complete picture of the environment to the user.
Autonomous Vehicles
FPGA Implementation of Multi-beam Beamformer and RaDaR Processor for Highway Pothole Detection and Avoidance
AP015 »
Automotive RaDaRs that incorporate ADAS-AD perform a wide spectrum of monitoring tasks for event-free autonomous navigation. Most such systems are pulsed RaDaRs, because this principle provides high dynamic gain, short measurement times and unproblematic signal processing. On a subjective level, such ADAS application demands an antenna radiation pattern tailor-made for its efficient functionality. Having electronic beam steering capabilities with good resolution and imaging properties, in addition to satisfying the radiation pattern morphologies and peak side lobe levels, would give more functionality to existing ADAS applications.
Potholes on interstate highways and local roads have been a bane of road transportation worldwide. The presence of potholes has a devastating impact on the vehicular platforms – with the attended fiscal element – in addition to hampering a smooth and predictable flow of traffic that is paramount to future autonomous road transportation. The impact on air quality and driveability, on account of the sluggish traffic flow over a stretch of such degraded tarmac generating harmful vehicular exhaust from the laboured performance by its engines, is substantial.
A medium range RaDaR (MRR) would serve as a capable ADAS tool in this context. An MRR sets a requirement for availability of multiple instances of narrower beams that shall enable tandem detection of as many potholes over a particular stretch of degraded tarmac. An added functionality in this operation is the processing of RaDaR 2-D data in real time. The high spatial resolution radio image data would offer a measure of the pothole parameters, against which the autonomous driving aspect could be altered by a suitable proportion in the interest of defining the traffic predictability criterion. Such operations are highly computation-intense, with the need for agile engines to be implemented in hardware and software. The inherent parallelism and reconfigurability features offered by FPGAs would ably assist by providing an opportunity to specialize the data processing implementation as needed. This opportunity is likely to grow as the logic capacity of FPGAs increases and the cost of FPGA devices is reduced. In the aforementioned context, FPGA-based beamforming and radar processing systems would be developed, tested. and deployed. The RaDaR would be designed as capable of deploying upto 10 highly-directive beams that could enable sensing over a range of 50 m, with a range resolution of few tens of cm and a spatial resolution of few cm, without the aid of external delay elements. The use of FPGAs in vehicular radar systems and an analysis of their robustness would be comprehensively evaluated during the project.
An added element in the work shall be the geo-tagging aspect, that would help generate a vast database of all such pothole detections across a span of geography criss-crossed by all types of roads. A low-latency communication infrastructure would be paramount to this operation, given the pervasive access requirements placed on the data by multitudes of vehicular systems before they encounter any pothole from the catalogue. The database requirements could be met if they are cloud-based systems, that feature elements of data and system security.
Health
IOT BASED ELECTROLYTE BOTTLE
AP017 »
During recent years, due to the technological advancements many sophisticated techniques has been evolved for assuring fast recovery of the patients in hospitals. Need for good patient care in hospitals, assessment and management of fluid and electrolyte is the most fundamental thing required. Almost in all hospital, a nurse is mainly responsible for manual monitoring of the electrolyte bottle level as most Hospital uses simple electrolytes bottles with no indication and electrolyte is the most fundamental thing required in hospitals and unfortunately most of the time, the observer may forget to change the bottle at correct time due their busy schedule due to which reverse blood flow starts from patient to electrolyte bottle due to which it can cause death of patient.
To overcome this critical situation, a IOT based automatic alerting and indicating device with portable cover is proposed.
We will design a portable cover system for such electrolyte bottle. Those Wearable cover will have non contactless water level sensor
along with GPS system on sides of bottle and non contactless water level sensor will detect level of fluid inside bottles. Such data will also send to nurses and/or doctor`s mobile through GPS system and they can start or stop the fluid and also will have monitoring fluid condition with security password also.
This project is completely IOT based which will break the wall of traditional methods used in Hospital and help in more fast and securing way and helps in fast recovering of the patients in case of emergency and prevent the panic situation in the hospital causing during manual monitoring of Electrolyte Bottle and will be a great assisting to the nurses which will reduce their stress and will be a great help to their work and and will focus on hospitals to equip with advanced Technology.
In, Hospital ICU, CCU, NICU, OPD, OT, most of all department of hospital required such kind of automatic monitoring and indication system. Also Health care industries will one of the users. such monitoring system can be useful in small , medium and large size of hospitals and also it useful during home care. This device will decrease the chances of patients hazards and increases the accuracy of health care in hospital. This is of high advantage to the patients especially during night times. This system also avoids the fatal risk of air bubbles entering the patient’s bloodstream, which is a serious threat as air bubbles in blood can cause immediate death. Such a device will create assurity of non-harm condition to patients and also helpful to monitoring of data and such data can be stored and will be useful in future. This project will be a great help in preventing hospital nurses to avoid direct contact with patients in hospital.
Smart City
A design for future city
AP018 »
A city with IoT connectivity, autonomous car is a city of the future.
This project will try to accelerate network architecture search algorithm on FPGA which will usually take hours and even days on decent GPU.
Some of the project modules which are currently being worked on :
https://github.com/promach/gdas
https://github.com/promach/DDR
Data Management
拥抱未来
PR010 »
电子工程系参赛队伍
Transportation
eego
AP021 »
With the world moving towards electric vehicles to combat the problem of emissions from vehicles, this project intends to build a battery based autonomous bot by adding AI to the bot with an emphasis on solar power along with battery management. Though there are a lot of organizations trying to contribute or gain value in the autonomous domain there is no single open project to understand and contribute to autonomous bots. This project intends to fill the gap by building a bot with AI and smart energy capabilities. The first level scope is to make the bot navigate by itself between two points in a defined environment i.e. L3 autonomy level and charge itself via solar power. This involves various sensor inputs for sensor fusion; wheel encoder for speed control, Object recognition for obstacles using camera, LiDAR for depth , RADAR for trajectory tracking and a powerful edge AI device for compute. The PID feedback control for the drive is difficult to achieve via software, so this will be developed on FPGA as well as any hardware acceleration ( RADAR data processing for instance) required for sensor fusion. Efficient solar power and battery management will also be targeted. Unknown obstacles will be reported to the cloud where models are trained ( by other bots as well in larger scope) and new models will be reloaded into the bot for better AI capabilities. Note: ROS is tentative for the sensor control, PID on FPGA for drive control, RADAR usage might be out of the scope based on availability of the sensor.
Smart City
Smart traffic light Control system Using FPGA
AP022 »
Now a days, controlling the traffic becomes major issue because of rapid
increase in automobiles and also because of large time delays between
traffic lights.
So, in order to rectify this problem, we will go for density based traffic
control system.