InnovateFPGA Design Contest Winners

Drone Package Delivery Safety

Foale Aerospace Inc.

Problem:

Package delivery is a vital component of a smart city infrastructure; however, it comes at the cost of traffic congestion and Carbon Dioxide emissions, a major contributor to Global Warming. By 2050, last mile package delivery could produce more than 2 million tons of CO2 per year.

Solution:

This project seeks to develop an aerial package drone delivery system that is renewably powered and can replace last minute point to point courier delivery in cities that rely on CO2 emitting services today. The project can detect and communicate strong or abrupt changes in air motions as well as create a “digital twin” for development. This promotes safety for the drone as well as for nearby people and property.

Project Details:

The reprogrammable DSP development capabilities of the FPGA enables the drone to react to 8 concurrent analog channels that measure the rotation and acceleration of the drone. The HPS provides delivery of hazardous event telemetry to Microsoft Azure IoT hub and added a ‘digital twin’ capability that can reproduce real events for algorithm development using the Analog Devices DC2025A-A DAC.

Cloud-Based Management of Solar Converters

Daniel Chavez (Universidad Nacional de Ingenieria)

Problem:

As more renewable power generation capabilities, such as solar and wind, become more widespread, distributing the power load across the electrical grid becomes more complex. There are frequent mismatches between power demands and generation that need to be managed.

Solution:

This system is designed to collect local point of use energy data more efficiently from many sources that could be used to control electricity generation. This data is reported to cloud-based systems which could then be combined with a grid policy to better optimize power demand and generation.

Project Details:

The FPGA directly measures and controls the modulation of a single-stage bidirectional DABRS converter and wirelessly communicates the data to Cloud-based Azure server MariaDB database. Since the data is processed locally by the FPGA, it minimizes the telemetry needed to communicate back to cloud-based central control.

Fruit Waste Reduction System

Nixon Fernando Ortiz De La Cruz (Universidad Nacional de Ingenieria and Nacional Mayor de San Marcos)

Problem:

Smaller farming enterprises often lack the resources to minimize waste during the distribution and transportation of their produce. Due to a lack of high-quality storage and transportation systems, a large percentage of produce is spoiled prior to eventual delivery to consumers. 80% of mango fruit production in Peru comes from small family-owned farms.

Solution:

This smart system will monitor, report, and manage storage and transportation conditions to reduce the amount of spoilage.

Project Details:

The FPGA aggregates data from a range of sensors. It also controls actuators that change air condition (i.e. CO2) and temperatures to reduce pre-mature ripening of the fruit. The video images from the camera are pre-processed by the FPGA, then sent to the cloud to analyze color data to correlate to fruit ripeness. This diverse dataset is sent Microsoft Azure IoT Hub for storage, processing, prediction and control of the smart fruit storage/transportation containers. Since most of the processing is done by the FPGA, it minimizes data sent to the cloud, reducing cost of transmitting large amounts of data.

Intelligent Farming Optimizer

Jyotsna Bavisetti (Rajiv Gandhi University of Knowledge Technologies, Nuzvid)

Problem:

Choice of crop selection by farmers can often be uninformed and inappropriate when using traditional farming methods. This leads to lower yields and quality of crops because of a poor understanding of water and soil amendments. Food supply in developing countries is especially vulnerable.

Solution:

This comprehensive farmer aid can recommend suitable crops based on the soil condition, climate, and water availability of that region. It can control and optimize irrigation, detect plant disease, detect weeds, and provide farmer guidance.

Project Details:

The FPGA aggregates data from a camera and wide range of Analog Devices sensors. It is capable of measuring nitrogen, phosphorus, potassium, pH, water level, soil moisture, and temperature. This diverse dataset is sent to the Microsoft Azure IoT Hub for storage, analysis, display and farmer guidance. A machine learning algorithm predicts the most appropriate crop for the conditions and controls irrigation. An object detection model can identify weeds, and an AI-based algorithm was trained to identify disease to recommend treatment.

Consumer Mini-Greenhouse System

Pahan Mendis (University of Moratuwa, Sri Lanka)

Problem:

There is growing awareness by consumers that certain large-scale farming and transportation of produce can have a negative effect on the environment. There is a rising desire for consumers to grow their own food not only for environmental concerns, but also to improve their quality and food safety. However, many of these aspiring farmers lack the knowledge or capability to do so efficiently.

Solution:

This smart and automatic mini-greenhouse management system targets food production for the urban community consumer who may have minimal farming expertise. It can provide guidance for irrigation, fertilization, ventilation, and light.

Included AI features can predict yield and identify abnormal growth behavior.

Project Details:

The system captures data from a camera and wide range of Analog Devices sensors (measuring air humidity and temperature, soil pH/moisture/temperature, CO2 levels, and light intensity). In parallel to the sensor data, the FPGA accelerates CNN AI-based image processing. This diverse dataset is sent Microsoft Azure IoT Hub for storage, processing, and prediction of results over time.

Mental Health Advisor

Sudhamshu B N (Dayananda Sagar College of Engineering)

Problem:

Mental and behavioral cases make up an increasing percentage of worldwide health problems. However, such cases remain grossly under-represented in conventional public health statistics (which focuses on mortality vs. other factors, such as dysfunction).

Solution:

A smart glove collects various human body and environmental parameters for machine learning models to analyze and classify symptoms of various targeted mental health conditions. Based on user health and mental state, timely positive suggestions are communicated back to the consumer as “recommendations”. Mental health is given equal importance to physical health.

Project Details:

The FPGA aggregates data such as air and body temperature, sweat gland activity, light conditions, and air quality from sensors on a smart glove. It then uses machine learning to characterize and correlate data with mental conditions. Live video of the patient from a camera input, feeds into a video emotion recognition model to correlate data into 5 traits (Angry, Anxiety, Happy, Neutral, and Sad). This diverse dataset is sent to the Microsoft Azure IoT Hub for storage and machine learning processing. Recommendations are sent back to the user’s mobile phone app.

Coral Reef Recovery

Jose Filho (King Abdullah University of Science and Technology)

Problem:

25% of all marine life depend on healthy coral reefs, however, increasing ocean temperatures are causing some corals to bleach. The bleaching process occurs when the coral expels algae living in its tissues. This algae is critical to the survival of the coral.

Solution:

Laboratory studies show that certain beneficial microorganisms for coral (BMCs) can stop the bleaching process and allow the coral to recover. This system can deliver the coral probiotics in the marine environment and monitor its long-term efficacy.

Project Details:

FPGAs allow for a flexible and reconfigurable experimental platform. The FPGA gathers data from cameras, temperature sensors and sea luminosity data from an Analog Devices Ultralow Power Light Recognition System. The FPGA uses AI to accurately determine the bleaching stage and make a fast decision to deploy the BMC. This solar powered experiment is deployed near shore and is able send 4G data to Microsoft Azure to visualize and manage the rehabilitation process.

Indoor Air Quality Management

Ricardo Núñez Prieto (NVISION s.l. / Institute of Microelectronics of Barcelona (CSIC) / UAB)

Problem:

Evidence has linked chronic exposure of CO2 concentrations as low as 1000ppm to several human health disorders. Studies have also demonstrated that viruses are released during exhalation, talking, and coughing, and that transmission is more likely to happen in indoor environments.

Solution:

Measuring exhaled CO2 is the best available low-cost method to assess air quality risks. This project can derive CO2 concentration and send actionable information to the Microsoft Azure server to manage alerts and ventilation.

Project Details:

This system uses data from an Analog Devices NDIR CO2 sensor system that can measure CO2 concentration in the range of 400-5000 ppm as well as a temperature sensor. An FPGA is a good choice for this application as it has the flexibility to support multiple sensors and can also be adapted to other sensor types, algorithms, and communication protocols. The project is optimized for a minimal system energy footprint by running the CO2 concentration algorithm locally in the FPGA to minimize the communication of complex data.

Smart Farm Control System

Mohamed Abdelaziz Louhab (University M'hamed Bougara Boumerdes)

Problem:

Minimizing cultivation mistakes is important to minimize costs as well as ensuring food security in regions that do not have optimal growing conditions. Other risks such as fire, animal encroachment, or theft, can have a significant consequence to a farm or a community.

Solution:

This smart greenhouse incorporates sensors that can monitor plant health using environmental factors such as temperature, humidity, and gasses such as O2 and CO2. The system can not only provide farmer guidance but also control irrigation, heating, and cooling.

Project Details:

The FPGA is the solar-powered brain of the system, using the HPS to implement a solar tracking algorithm to maximize output from the PV panels. The greenhouse is designed to minimize cultivation mistakes and make the produce tastier using 6 different Analog Devices sensors to monitor and maintain crop health. The system mitigates concerns of crop loss by using IR and PIR to detect animal presence and is also capable of fire detection and suppression.

Automatic Garbage Sorter

Longfei Yang (Hubei University)

Problem:

Global garbage production increases every year, polluting soil and water sources. Landfills can contain hazardous chemicals or other harmful substances that can enter the ecosystem causing harm to humans and the environment.

Solution:

An automatic garbage sorter can help reduce environmental pollution, save land resources, and promote resource recycling. This system is designed to reduce the thoughtless disposal of recyclable garbage and the classification of waste maximizes the potential to re-use waste and reduce harm to the local ecology.

Project Details:

This project uses image recognition to classify waste into 4 categories: Recyclable, Biodegradable, Hazardous and Other. An infra-red sensor detects the disposal of a new object, causing a camera to send an image to the FPGA. The FPGA implements a deep Convolutional Neural Network (CNN) called VGG-16 for the image recognition. Recognition and classification are enhanced with an OpenCL-based FPGA accelerator called pipeCNN. The garbage sorter moves waste into the appropriate bin with 95% accuracy and has a recognition time of 1.93s.

Pavement Damage Detection System

Dingwei Chen (Chongqing University)

Problem:

Pavement damage on roads not only affects appearance and driving comfort, but if maintenance needs are not identified and addressed, the road surface and associated structures will deteriorate which could lead to accidents or even loss of life. Road construction with concrete or asphalt both generate harmful emissions and pollution.

Solution:

This automated detection, location and reporting system uses 3D lidar and cameras to capture real-time road condition information that can be analyzed to determine if repairs are needed. Accurate inspection information coupled with a precise location allows maintenance to be efficiently managed and avoid costly road surface deterioration.

Project Details:

The intelligent road damage detection system uses lidar and a camera with an IMU (Inertial Measurement Unit), to obtain point cloud data and image information of the road surface. The system can correct for point cloud motion distortion and is able to synchronize the lidar distance information to the camera pixel information. This data can be used by cloud applications to combine damage details (size/shape of defect) with precise map data. Maintenance personnel are then able to focus on scheduling timely repairs.