Arduino Day Workshop: NanoDrone II: AI and Computer Vision with LoRa (Win a PSoC6 and a Pair of MKR 1300 Boards!)

re: Arduino Day Workshop: NanoDrone II: AI and Computer Vision with LoRa (Win a PSoC6 and a Pair of MKR 1300 Boards!) by maxpowerr

maxpowerr — Fri, 05 Mar 2021 12:39:28 GMT

I am gradually pursuing the development of artificial intelligence and computer vision. But I never owned quadcopters, so it would be interesting to try to assemble a quadrocopter with AI. The topic is very interesting and I will look forward to it.

Enrico Miglino well done!

re: Arduino Day Workshop: NanoDrone II: AI and Computer Vision with LoRa (Win a PSoC6 and a Pair of MKR 1300 Boards!) by ilvyanyatka

ilvyanyatka — Sun, 07 Mar 2021 06:16:52 GMT

Congratulations with the new book!

In a couple years I plan to move to a farm. And the drone I want to build will be a "Sheppard" drone. The idea is - (and I am not sure how doable it is, just an idea) I can code the path to fly and it will fly around the border of the farm. It will scan the view for sheep and goats (using image recognition - sheep is a moving white ellipse, goat is a moving gray ellipse). It will transmit video to my receiver, so I can monitor it. And it will notify me with a special sound signal if he recognizes a sheep or a goat near the border of the farm. So I can go get it back and the animal will not be lost in the woods.

re: Arduino Day Workshop: NanoDrone II: AI and Computer Vision with LoRa (Win a PSoC6 and a Pair of MKR 1300 Boards!) by maxpowerr

maxpowerr — Sun, 07 Mar 2021 14:16:49 GMT

I got a few ideas for farming:

1) Scanning and mapping a plot of land to find stress-prone areas that require attention (additional moisture, fertilization and pest spraying). This can be useful for sowing large areas of land.

2) Inspection and scanning of fruit trees for pests and disputed fruits.

3) Inspection and scanning and compilation of a map of forest areas for the presence of pests. I myself ran into this problem when small areas of the forest simply dried up due to the presence of pests (bark beetle).

re: Arduino Day Workshop: NanoDrone II: AI and Computer Vision with LoRa (Win a PSoC6 and a Pair of MKR 1300 Boards!) by jasperchua99

jasperchua99 — Tue, 09 Mar 2021 12:51:00 GMT

I am doing a project on drone AI and computer vision where im comparing different hardwares to see which one is able to allow the drone react faster to commands typed from the ground station.

I am hoping to build a system that allows integration of AI for commercial and military applications using modular payloads and recognition based on human facials, barcodes and even remote parcel collection at different terrains to allow better accessibility for payload transportation

The system can be built in such a way that different programs and payloads allows modularity and different types of application for the commericial and military use based on a set system configured to the clients need.

Currently, The existing raspberry pi with raspberry pi camera system mounted to my drone has a similar system but uses opencv and ros to operate the vision system.ill like to explore with tensorflow to further develop the recognition capabilities so that its able to improve on its own on the different conditions of the situation so that the next will allow better flying trajectories to reach the intended target. But, i am experiencing issues of not able to fall back to a back up system if the recognition fails. So i hope thats where the arduino can help reset the system midair

This is the Test of the drone using OpenCV, ROS with Raspberry Pi 3 and camera only.
Testing of Autonomous mode using Raspberry Pi

re: Arduino Day Workshop: NanoDrone II: AI and Computer Vision with LoRa (Win a PSoC6 and a Pair of MKR 1300 Boards!) by ralphjy

ralphjy — Tue, 09 Mar 2021 17:01:37 GMT

I've got a barebones Alphabot chassis that seems like it would be ideal for implementing an AI enabled rover bot. I bought the chassis when Seeed was having one of their 70% off sales .

The chassis is set up to use an Arduino Uno+ for control with a Bluetooth shield for communication and has a provision to add a Raspberry Pi for AI capability. I'd love to get one of these drone kits and adapt this chassis to create a rig with capability like Enrico did. It seems like I could implement all of the control functionality and AI functionality with some rewiring and 3D printing, and at least I'd have a head start with the chassis, wheels and motors and drivers. I'd also need to add a pan/tilt module for the camera.

Here's what a normal fully configured Alphabot looks like:

Here's what the section for mounting the Uno+ board looks like:

I could build a PCB to use the Nano Sense 3 and LoRa boards in place of the Uno+.

I'd like to use the RPi with facial recognition to create a person following rover. A possible iteration is to add a GPS to make the rover more autonomous. The GPS could provide coarse location and the camera could then be used to identify waypoints for guidance.

re: Arduino Day Workshop: NanoDrone II: AI and Computer Vision with LoRa (Win a PSoC6 and a Pair of MKR 1300 Boards!) by alrms

alrms — Wed, 10 Mar 2021 16:40:53 GMT

The drone I'd like to create is a co-drone or collaboration drone. This drone will be a companion to my everyday tasks at home.

The capabilities I'm thinking it should have can recognize me or any human it's accompanying and be able to approach my arm when I stretch it and press it's built-in health sensors down to collect my vital sign. It can act as a companion drone flying by your side as you move from one place to another. The drone can also be instructed to move through your gesture, telling it to go near or far from you and then task to collect environment data such as lighting, noise, temperature and humidity. Small tools and parts can be carried by the co-drone to support in your work as you move. This is most useful when you’re picking a nice spot to do a picnic or placing an object or plant when you farm. These data are relayed to your smartphone that you can view even if the drone is from a far distance. It can also create geotags using GPS and Lora to make it area monitoring easier to track and view

With the training, I hope to further enhance the drone's capability for emergencies such as reporting incidents and doing remote monitoring in our warehouse to check for any structural issues.

re: Arduino Day Workshop: NanoDrone II: AI and Computer Vision with LoRa (Win a PSoC6 and a Pair of MKR 1300 Boards!) by cgm11

cgm11 — Wed, 10 Mar 2021 17:59:59 GMT

Hi there,

I am engaged in a project to protect Jaguars here in Argentina (they are under risk of extinction, only 250 alive today and they are important to keep the balance of the environment, without them, harmful species for forests grow without any control thus they deteriorates trees, thus birds are gone, then no one spreads seeds, then you can imagine),

I published a few things here (tinyML for Cow Collars to detect anomalies, retaliation killing is one of the most frequent death causes for Jaguars here) but as my new devices are based on the Sony Spresense I am not publishing too much here. Now, I started to incorporate the Arduinos Portenta H7 + VisionShield Lora, you can see some details about my project here: www.wildedge.info (I am going to update the site in a few days).

Drones are a strategic thing to me, as you may imagine, deploying devices on wild places means that maintaining those devices could be costly, specially because they are deployed in hard to access places. The role I see for drones (the technology being proposed is quite similar) is to have an automated mechanism to update devices, retrieve data and clean them. I Use LoRa, GPS and BLE. GPS is used to locate the devices, LoRa is there to change device status / alert modes and monitor the activity to early detect dangerous situations for Jaguars. In the context of the automated update, LoRa will be used to alert a device that it is going to have visits!!! The drone must fly autonomously to the specified devices, get close, stay steady, perform a BLE handshake and then retrieve data, clean, push updates (firmware, configuration and models) and done, gone to the next one. A wireless charging mechanism should be ideal however weight of batteries and costs could be complex to handle. When things get stable and trustable having drones to inspect zones would be quite valuable, however, 100% autonomous drones near people is still banned here in Argentina.

I hope I get your support, it is not easy to run this project.

Cheers,
Carlos.

re: Arduino Day Workshop: NanoDrone II: AI and Computer Vision with LoRa (Win a PSoC6 and a Pair of MKR 1300 Boards!) by justjeff

justjeff — Fri, 12 Mar 2021 16:15:19 GMT

Looking at this to exam wildlife types with vision and terrain information.

re: Arduino Day Workshop: NanoDrone II: AI and Computer Vision with LoRa (Win a PSoC6 and a Pair of MKR 1300 Boards!) by ami301618

ami301618 — Fri, 12 Mar 2021 16:16:46 GMT

I am working on a project estimating the ripening of fruits and take the decision to pluck them. It requires AI and computer vision for better results. As in hilly areas, it becomes very difficult for farm people to monitor and plan for the cutting of fruits. This drone which will be designed at a low cost will be more beneficial for the farm people and prevent fruits from over-ripening. I hope I could get your support in designing and developing low-cost drones for farming in hilly areas.

Thanks and Regards.

re: Arduino Day Workshop: NanoDrone II: AI and Computer Vision with LoRa (Win a PSoC6 and a Pair of MKR 1300 Boards!) by mekkatronix

mekkatronix — Fri, 12 Mar 2021 19:54:44 GMT

This discussion is very interesting to me as I am recently introduced to TinyML, Tensor Flow Lite, and other aspects of AI. I would like to develop a low-cost UAV that utilizes an Arduino Nano Sense 33 board along with computer vision, with the goal of identifying joints and discontinuities (cracks) within rock slopes and ultimately classifying a rock slope with regard to stability (i.e., the slope appears to be generally stable, or it appears to represent an unstable condition that is susceptible to rock fall, etc.).

I hope to learn more from this Workshop as well as from the community which surrounds it.

Many thanks for all of the contributions.

re: Arduino Day Workshop: NanoDrone II: AI and Computer Vision with LoRa (Win a PSoC6 and a Pair of MKR 1300 Boards!) by aniketdhole

aniketdhole — Fri, 12 Mar 2021 19:58:30 GMT

Hello Everyone,

I am working on Project to Detect Thefts and Riots using CCTV Cameras and Raspberry Pi and Intel Stick.

The main purpose of this drone is to survey a huge area from one place without needing the security person to physical monitor it. It will discover, detect, observe and interdict unusual events to avoid Crimes/Thefts.

I have an idea to make using this on Drone for patrolling area using

1. Raspberry Pi 4 (Already Have)

2. Intel Neural Compute Stick (Already Have)

3. Pi Noir Camera - For Night Vision

4. Adafruit GPS HAT - To Locate the Drone

5. Arduino MKR 1300 - To Communicate with Control Centre

Idea Plan:

1. As I wanted to make the Computer Vision Faster on Edge I am planning to use Intel Neural Compute Stick on Raspberry Pi using Tensorflow and OpenVINO boosts the processing performance by 2-3x times.

2. The Pi Noir Camera Module is used as it supports Night Vision and is compatible with Raspberry Pi

3. The GPS Hat is for Locating and Positioning the Drone along with its own Flight Controller and to create a map of drone surveillance path.

4. Arduino MKR 1300 LoRAWAN is used to communicate alerts or events with control centre. As it doesn't supports High Bandwidth only limited data will be send in 500 metre range.

5. All the Computer Vision will be performed directly on Edge

I would also include PSoC6 for Drone Control after learning about it from Training.

With the Kit and Training I will be able to complete this project a lot easier.

Thank You.

re: Arduino Day Workshop: NanoDrone II: AI and Computer Vision with LoRa (Win a PSoC6 and a Pair of MKR 1300 Boards!) by kuatous

kuatous — Fri, 12 Mar 2021 20:15:38 GMT

This project seems amazing! My project goal is to utilize a computing hive cluster with a number of these drone analogs to produce an all in one mobile cultivation probe. Using a number of variable drones to collect necessary information to allow for mobile growing and monitoring.

re: Arduino Day Workshop: NanoDrone II: AI and Computer Vision with LoRa (Win a PSoC6 and a Pair of MKR 1300 Boards!) by dldatwyler

dldatwyler — Fri, 12 Mar 2021 20:55:29 GMT

I want to build an intermediate level dtone

re: Arduino Day Workshop: NanoDrone II: AI and Computer Vision with LoRa (Win a PSoC6 and a Pair of MKR 1300 Boards!) by etrnllrnr

etrnllrnr — Fri, 12 Mar 2021 21:14:14 GMT

Im looking to build a farm drone to detect if plants have any pests, weeds, insects

Looking to build a collection of coordinating drones to divide the farm & communicate with each other

re: Arduino Day Workshop: NanoDrone II: AI and Computer Vision with LoRa (Win a PSoC6 and a Pair of MKR 1300 Boards!) by moi8765

moi8765 — Sat, 13 Mar 2021 12:40:10 GMT

More than a project, my project proposal is more a part of the prototype development of a social entrepreneurship venture which I am pursuing along with my batchmates. We call our proposed venture ResQSens since it is focussed on drone-based disaster relief and recovery.

ResQSens provides an autonomous drone ecosystem for aerial surveillance, information dissemination, damage assessment and delivery of aid material to reduce the time needed for aid to reach the remotest of places struck by natural disasters. We aim to build financial and social capital by engaging with public authorities by helping them assess damage, speed-up relief efforts, and equip health-infrastructure.

We identified problems that affect resident communities and disaster management authorities and identified tangible and addressable issues some of them being -

A large amount of time for relief material to reach the affected communities at the remotest of places due to lack of information about damage
Lack of reliable information dissemination methods for rescue operations which creates a state of panic in the affected communities leading to more casualties
A large amount of time taken to clear debris and for impact assessment in turn leading to delay in rehabilitation. The primary reason for this is the absence of a mechanism to quantify damage and provide information to the management without manual surveying.
Lack of adequate health infrastructure, which is especially important in a post-disaster scenario. This problem is worsened by the disruption in road networks.

Solution

Our solution is a multi-pronged approach which tackles problems during and after a natural disaster and also provides long term solutions for the remote regions of the country which do not have access to adequate health infrastructure. For this, we propose to deploy unmanned drones which -

● Broadcast rescue operation information in the remotest of places in local dialect while surveying the affected area

● Deliver relief material to places where aid, otherwise, takes too long to reach

● Carry out automated assessment of damage by analysis of aerial footage and send it over to disaster management authorities in a geotagged manner

● Deliver medical supplies at a hyperlocal level in emergency situations

Implementation

There are many startups who are working in the field of surveying and deliveries. However our USP is that our drones are unmanned which helps us reach out to the large population in need at the right time

We would also offer management authorities the automated impact assessment feature which helps speed up rehabilitation efforts, which is something no organisation has worked on yet.

Furthermore, the same fleet of drones can be repurposed to multiple use-cases in during disaster impact, immediately after natural disaster and general scenarios.

The solution can be scaled up to rural areas, vulnerable to natural disasters globally.

The primary component of our solution are the unmanned drones that would perform aerial surveys and aid delivery at the hyperlocal level. The drones are loaded and serviced at one of several service stations at the district level. The stations are manned by trained personnel for charging the drone in between trips and to load/unload the drone.

The rescue teams communicate with the service agents to publish whatever information needs to be sent to the residents in the local language. The message is loaded on to the drone via our service app that does a basic OTA update of the information and broadcasts it using speakers mounted on the drone.

For aerial surveys, the downward-facing camera captures images of the affected areas. The imagery data from the Raspberry Pi Noir camera is processed using an onboard processor based on Raspbery Pi 4 coupled with an Intel Neural Compute Stick 2 and the objects identified in the scene are sent along with the drone`s GPS coordinates acquired from the Adafruit GPS hat to the AWS IoT cloud. An inference engine instance running on AWS DynamoDB then analyses these data points and sends impact assessment data to management authorities who can visualise the data on our app. The way the data is sent to the cloud is that first the data from the host PSoC 6 Microcontroller is sent over Bluetooth to a mesh network constituted by all drones currently in flight. This mesh is connected to a gateway based on the ESP32 Wifi module that actually pushes data to the cloud.

For medical deliveries, our service agents are contacted. The service agents configure the drone OTA to pick up and deliver supplies to and from particular sites. All this creates an entire ecosystem to help reduce damage and to help disaster-struck areas bounce back faster.

End Result of Project

An autonomous drone capable of unmanned aerial deliveries, broadcasting information about rescue operations, surveying and impact assessment.

Complimenting the drone, would be an integrated app for configuring the drone over the air, getting geotagged impact assesment data from the cloud and for keeping track of rescue operations.

Beneficiaries of Solution

Residents of disaster affected regions, disaster management authorities and governing bodies are the major stakeholders that we intend to benefit.

The benefits for the affected communities are -

Aid is able to reach hyperlocal communities much faster
Rehabilitation is speeded up thus helping people bounce back faster
Access to better and more reliable healthcare facilities

The benefits for Disaster Rescue staff are -

They get geotagged impact assessment removing need for manual surveys
They get precise information about what amount of aid is needed in a particular place
Broadcasting messages about rescue operations helps reduce chaos

The Benefits for the Local authorities are-

Due to the automated impact assessment, funds and other resources for rehabilitation can be doled out faster
Debris management is much easier

re: Arduino Day Workshop: NanoDrone II: AI and Computer Vision with LoRa (Win a PSoC6 and a Pair of MKR 1300 Boards!) by ajayvishaal

ajayvishaal — Mon, 15 Mar 2021 18:34:10 GMT

Hi everyone!

I just wanna share few words about my projects which I am working on. Actually I'm building two drones, one is "Aquila - a mini drone" and another one is "Chronus - an autonomous drone".

Aquila - a mini drone:

This is a mini hexacopter consists of an Arduino MKR wifi 1010 and an MPU6050 IMU sensor. This drone can be controlled using our smartphones through wifi. I've designed a mobile app using Python and Kivy framework to control the drone using smartphone. I'm writing my own firmware for the drone using arduino and soon I'll make my code open source so that people can learn more about drones and the technology behind it. I wanna share some of the pictures of my drone here.

Kivy app interfacce of Mini Drone:

The above are some of the pictures of my drone. I'm building this drone as a phone as well as voice controlled drone. This kind of mini drones will help students to develop their knowledge about drones.

Chronus - an autonomous drone

This is my another drone project. This drone has a Jetson nano on it for autonomous navigation purpose. I'm deploying ROS for making this drone autonomous. I'll be intgerating lidar for indoor navigation and GPS for outdoor navigation. SLAM will be used for autonomous navigation. The image of my budding drone is here.

Soon I'll be integrating Jetson Nano with ROS in my Drone. And I'm very much interested in this webinar. And I hope this webinar will give me more information, which will be helpful for me to deploy in my drones.

Thanks and Regards,

Ajay Vishaal T.

re: Arduino Day Workshop: NanoDrone II: AI and Computer Vision with LoRa (Win a PSoC6 and a Pair of MKR 1300 Boards!) by cgm11

cgm11 — Mon, 15 Mar 2021 18:39:04 GMT

What I see here is that there is a lot of people interested in this topic and most of all having active projects. Is there any focused group or community here?

Cheers,

Carlos.

re: Arduino Day Workshop: NanoDrone II: AI and Computer Vision with LoRa (Win a PSoC6 and a Pair of MKR 1300 Boards!) by n.jaksic

n.jaksic — Mon, 15 Mar 2021 20:20:31 GMT

We are developing a large octocopter drone capable of caring a 100Kg payload (maybe a human) as a part of search-and-rescue operations. We have most of the hardware and most of the preliminary designs but we didn't build it yet. We are still looking for the best controller and AI bells and whistles.

re: Arduino Day Workshop: NanoDrone II: AI and Computer Vision with LoRa (Win a PSoC6 and a Pair of MKR 1300 Boards!) by dramoz

dramoz — Mon, 15 Mar 2021 20:42:01 GMT

Two projects:

Survillance drone for my auto-lawnmower: direct the lawnmower to the proper place and monitor activity

In-house cleaning drone: this is in the long term, I am going to explore to use a drone to clean top corners. Later, check if a drone can replace a bulb.

Regards,

Danilo.

re: Arduino Day Workshop: NanoDrone II: AI and Computer Vision with LoRa (Win a PSoC6 and a Pair of MKR 1300 Boards!) by hauris

hauris — Mon, 15 Mar 2021 21:39:23 GMT

I would like to see a project that does "VIO", i.e. Visual Inertial Odometry (also called "VINS" = Visual Inertial Navigation System). This uses visual data (to determine pose) and inertial data (from the IMU accelerometer/gyro to also determine pose) and fuses the data using a Kalman Filter (MSCEKF = Multi-Constraint Extended Kalman Filter). There are other methods to fuse also. The fusing of the 2 data improves performance a lot. This would be really cool! Jon