In 2019, 35.704 accidents with victims on Portuguese roads were reported, of which 626 resulted in fatal victims, 2.168 seriously injured and 43.183 with light injuries.
Since its creation, at the end of 2018, there has been a strong growth in the use of technologies using 5G network, since it allows to send data at high speed.
Therefore, the exponential growth of intelligent solutions for the most varied activities, as well as the lack of technology in the market related to 5G, leads to believing that there is a good opportunity to create a solution to improve road safety, benefiting from the advantages of using this technology.
It is necessary to develop a service whose objective is to support the traffic control entities on roads, in order to have access to information and data processed practically in real-time.
It is intended to create a intelligent solution that takes advantage of traffic cameras and radars from the PASMO project, so that in the cloud computing environments from the 5GASP and the 5GAIner projects, it is possible to detect anomalous and relevant situations in the traffic flow of the A25, Praia da Barra and Costa-Nova.
It is also intended to implement a Car 2 Infrastructure / Infrastructure 2 Car communication using 5G, so that it is possible to exchange important information between vehicles.
Among the most varied objectives to improve the detection of different situations on the roads, the application must provide a dashboard with information about:
-
Number of vehicles to circulate (as well their velocity) at the moment and monthly
-
Traffic
-
Presence of animals, people or strange objects on the road
-
See the number of cyclists, runners or people walking on bike lanes
-
Detect inflow of the zone - with cameras estimating how many people are in Praia da Barra and Costa Nova. It also serves to tell if there are still parking spaces
-
Presence of stopped vehicles on the road
Nice to Have
-
Presence of vehicles on the side of the road
-
The velocity on the radar can tell whether the road and the weather are in good conditions for the current velocity. If people go too fast the conditions are good
Future Work
- Detect models of cars and check ambiental footprint taking in account max velocity allowed and model consumption
-
Detect adverse conditions
-
Rain sensors
-
Light sensors detect that the headlights turned on
-
Detect if the fog headlights turned on
-
Calculate the carbon footprint by detecting the model of the car and how much it consumes
-
Warn cars if another one is coming their way with excessive speed. This can be important in terms of vehicles with priority
Nice to Have
- Check if there are gas stations in the same road within a plausible distance, to know if with the current amount of gas/battery, a car can safely reach it. If a substantial number of cars pass with low amounts of gas and a gas station isn’t near, it might be a sign that there is a need to create a new one
Future Work
- Explore additional features with this type of communication
On the other end, it is important to develop a solution that is able to turn on or turn off cameras/servers programmatically, in order to manage the available resources in a better way possible. This will reduce the cost of systems renting, as a cloud service that makes available the VM´s that are used for data processing. Overall, the app may learn in which situations are necessary to change the VM´s state, e.g, if the camera goes down, the associated VM is turned off; If the camera becomes available again, the associated VM is restarted.
Car Communication (Hugo Almeida e Orlando Macedo)
-
Research OBDII protocol
-
Implement a car emulator to receive and generate data
VNF Orchestration (Hugo Almeida e Miguel Almeida)
-
Research about VNF in 5G
-
Integrate developed VNF into 5GASP testbed
Computer Vision (Miguel Almeida, Orlando Macedo e Carolina Araújo)
-
Research for Machine Learning and pattern recognition tools
-
Train a custom Object Detection Model
Backend (Carolina Araújo e Hugo Almeida )
-
Define what database to use for each scenario
-
Create the databases schema
-
Define the endpoints necessary to the API
Frontend (Carolina Araújo e Hugo Almeida)
- Design and develop an interface for the web applications
CI Pipeline (Orlando Macedo e Miguel Almeida)
- CI Pipeline
The goal of this project is to develop a set of VNF’s (Virtual Network Functions) that are capable of processing information from the video cameras and radars, notifying and storing data, using services and mechanisms made available by 5G.
It must analyze data in order to know the number of vehicles that are driving above the speed limit, the monthly average of vehicles, etc.
Besides that, it is imperative that the system is flexible, i.e, it should be able to manage the number of cameras and radars being used, detecting when they stop streaming data.
With this system, the users will be able to use the provided data to:
-
Decrease the costs of renting the VMs
-
Learn when it’s in their best interest to increase/decrease the number of lanes
-
Analyze statistics
With the communication between cars, it will be possible to determine the road conditions as well as to warn drivers in case of danger, such as with the arrival of vehicles at high speed.
Platform still under development based on existing inftra-structores that aims to shorten the idea-to-market process, offering a set of automated services, so that it is possible to develop and test new products built using 5G NFV architecture. 5GAPS offers software support tools, which cover use cases in different domains, in order to allow CI/CD of VNF´s in a secure and trustable environment.
The main objective of the PASMO project - Open Platform for the development and experimentation of Mobility Solutions is the design, implementation in the field and provision of a support platform for intelligent mobility, open to the participatory experimentation of companies that can effectively collaborate to test (technology ) and validate (market) equipment, protocols, processes, applications, standards and services.
PASMO project is hosted inside IT at Universidade de Aveiro, Portugal. Has as coordinator the Professor Joaquim de Castro Ferreira.
One of the projects associated with PASMO project is the dissertation of a student of UA, Rafael Oliveira. This dissertation is important in the way that has similar features to what should be implemented by us.
The motivation of Rafael was associated with the ambition of having a smart city. Regarding that, our colleague developed the following features:
-
Map with vehicle velocities
-
Dashboard with information about capacity at parking lots
-
Dashboard regarding average velocity, percentage of vehicles caught speeding, etc.
-
Information about radares and cameras, and their geographical location
-
API exposing all the available functionalities
The main objective of 5GinFIRE is to build and operate NFV (Network Functions Virtualization) as an ecosystem of several experimental installations allowing the instantiation of software over them. It was a 3-year innovation and research project supported by the European Union, starting in January 2017.
With the growing popularity of Cloud services like those from Amazon or Google, more and more applications are hosted on this type of services. In recent years, most telecommunications service providers have incorporated some concepts and Cloud technologies in order to use homogeneous hardware, to disassociate applications from the hardware where they are being used and to manage resources so that they are more efficient. Network Functions Virtualization (NFV) were built based on these characteristics, running Virtual Network Functions (VNFs) on top of homogeneous hardware. Examples of this are routers, firewalls, NAT services, etc., all of which are virtualized, on similar hardware, and managed through NFV.
Most Cloud infrastructures are centralized data centers, which is not ideal for applications with some requirements such as low latency. Using NFV technology, on many occasions, VNFs are executed in specific locations on the network with the use of 5G, allowing for better latency.
5GinFIRE, based on these conclusions, designed and operated multiple installations with 5G NFV in order to allow the execution of VNFs, the management and allocation of resources and the communication between the various facilities.
In China’s most recent plant, 2015-2020, it’s been estimated that investments of upwards of $400 billions will be made in the 5G field, creating at least 8 million new jobs.According to Deloitte, China is leading the race on 5G.
Smart highways are being created, supporting cellular network-coordinated transportation, that will avoid the necessity of having humans controlling the traffic. An intelligent system with prediction abilities based on the traffic information fed in real time is expected to be created to help autonomous vehicles. This system will rely on more than two thousand 5G base stations.
In Fangshan, the 5G technology will allow autonomous vehicles to communicate with each other in order to avoid collisions.
Another chinese company is creating the first 5G electric and autonomous buses.
Agile methodologies are applied, more specifically SCRUM, using Sprints of 2 weeks with rare exceptions:
Milestone 1 - presentation of the lifecycle objectives and calendar for the project
Tasks
- Overall Project Planning
Milestone 2 - presentation of the lifecycle architecture; the milestone is achieved when the architecture has been validated
Tasks
-
Car Communication #1
-
Computer Vision #1
-
Backend #1
Milestone 3 - prototype; mid-term presentation with supervisors; peer evaluation
Tasks
-
Car Communication #2
-
Computer Vision #2
-
Backend #2
-
CI Pipeline #1
-
Frontend #1
Milestone 3 - prototype; mid-term presentation with supervisors; peer evaluation
Tasks
-
Car Communication #3
-
Computer Vision #3
-
Backend #3
-
CI Pipeline #2
-
Frontend #2
-
VNF Orchestration #1
Milestone 3 - prototype; mid-term presentation with supervisors; peer evaluation
Tasks
-
Car Communication #4
-
Computer Vision #4
-
Backend #4
-
CI Pipeline #3
-
Frontend #3
-
VNF Orchestration #2
Milestone 4 - project presentation; all functionality has been developed
Tasks
-
Technical report #1
-
Polish overall project #1
Milestone 4 - project presentation; all functionality has been developed
Tasks
-
Technical report #2
-
Polish overall project #2
Milestone 4 project presentation; all functionality has been developed
Tasks
- Technical report #3
-
Team Communication - Discord was chosen since all team members are already familiar with the software.
-
Backlog Management - Github Kanban to record the backlog of tasks, being possible to associate each branch to an issue, track which issues are on-going, which have already been done or which are giving problems, for example. Each repository will have its issues, and there will be GitHub projects for each Sprint in the organization.
-
Repository - Github was chosen as the git platform to store the code based on an organization that aggregates all the project's repositories and documentation.
-
Git Workflow
The chosen workflow is based on the feature-branching with the following branches:
-
Master - Main branch, where a commit is made for each release.
-
Develop - Branch focused on development, where new branches for features come from and where they are at the end of their development.
-
Feature/* - One branch per feature. It originates from develop and is where it is merged when finished. To merge, a pull request must be made.
-
Release/* - This branch originates from develop and is generated at the end of each iteration, before its launch, for the final finishes. When the improvements are finished and the launch can be carried out, it is merged into the master.
-
Hotfix/* - This branch originates from the master and is opened when there is a need to correct an error identified in production and which cannot wait for the next release to be corrected.
-
Formal Meeting Platform - Zoom
-
File Sharing - Google Drive
-
Advisor - Prof. Diogo Gomes
-
Co-Advisor - Prof. Joaquim Ferreira
-
Co-Advisor - Prof. Rui L. Aguiar
-
Product Owner - Carolina Araújo
-
Software Architect - Hugo Almeida
-
Team Manager - Miguel Almeida
-
DevOps Master - Orlando Macedo
-
To collect the requirements, the first thing that the group did was try to understand better all the technologies involved. In specific, technologies related to 5G communication, real-time object detection and projects already made that could help us recognize better the functionalities that could be developed.
-
After collecting the knowledge about the domain, a brainstorm was made between the elements of the group. The major part of the ideas related to this project were made at this brainstorm.
-
At last, all the ideas/requirements were discussed with the advisor Diogo Gomes and co-advisor Joaquim Ferreira. The advisors selected in terms of easiness/risk of implementing these requirements, as well as whether they were feasible or not.
-
The VNF infrastructure that our project will need is still in development, so this is something that the groups needs to have in count, because if the development of this infrastructure don't go as planned, such situation will be very problematic.
-
5G Broker isn't fully implemented yet, one more time, the good evolution of our project is somewhat dependent on the completion of the 5G Broker.
-
There are legal risks associated with the fact that our project uses video streaming from public motorways, but all those risks were avoided meticulously.
At this section it will be presented a brief description of how the system is expected to be used by the stakeholders.
To start, vehicle drivers must have a OBU (on-board unit) that connects to the obd2 port of the vehicle. This connection allows the retrieval of information referent to the car. Lastly, this information is transmitted to our infrastructure, in order to be processed and provided on the web application, such service will help other drivers to drive more safely.
As it was said, the web application will have information associated with the traffic and can be accessed by everyone without the need to make login. The services provided by the web application will be detailed further up but succinctly we can talk about data referent to traffic at a specific zone, number of people cycling or running, and more.
With the objective of implementing a real-time object detection, several of algorithms were considered. Among them SSD (Single Shot Detector), R-CNN (Region-based Convolutional Neural Networks), and, what turned out to be chosen, YOLO (You Only Look Once) due to the perception that this was the fastest algorithm with the best results.
This statement is proven by the following article from the authorship of Ambika Choudhury.
According to the researchers at Facebook AI Research, the unified architecture of YOLO is extremely fast in manner. The base YOLO model processes images in real-time at 45 frames per second, while the smaller version of the network, Fast YOLO processes an astounding 155 frames per second while still achieving double the mAP of other real-time detectors. This algorithm outperforms the other detection methods, including DPM and R-CNN, when generalising from natural images to other domains like artwork.
With regard to car communication, it was important to investigate technologies that could collect information from cars and forward them to our infrastructure. In this way, it was very productive to understand better the projects associated with 5GinFire. In particular the SURROGATES experiment, this experiment used a OBU (on-board Unit) that forwarded the data of the car, that was being provided by the connection to the obd2 port from the car, to a RSU (Road Side Unit) that sent that data to a vOBU (virtual OBU), allocated on a remote location using VNF. So, this experiment was very important to understand what kind of technologies were being used and what we could also take advantage of.
The Car 2 Car consortium was also crucial to understand what kind of services are being developed in the present in regard to car communication. Some of the requirements found were inspired by what this consortium is trying to achieve. One of those examples, is precisely the alert of careless drivers to the rest of the motorway users.
"The basic principle is cooperation, i.e. the exchange of information between all road users, from road and rail vehicles to cyclists and pedestrians, and between road users and infrastructure elements such as traffic lights and road signs. All that happens in real time and at high speeds.” Fritz Kasslatter, Research Group industrial networks Austria
For more information, please read this article very insightful from Siemens.
To understand better if 5G technologies were being used in the highways, the same way that our group intended to do, a brief investigation was made, and we come across, among others, the article from Road Traffic.
In this article is said Chinese state-owned telecommunication company China Mobile is building a 5G smart highway project in Wuhan in Hubei province.
Their intention is to collect real-time traffic information, make AI-assisted forecasts using the data, which will support driverless cars. And this is possible due to the fact that 5G has low latency and allows communication between devices at real tistrong textme.
5GASP is a project developed in IT of Aveiro that will allow the group to test the VNF's created, making sure of CI/CD.
PASMO project was implemented with the intention of making the city of Aveiro a lot smarter. With the use of radars, cameras that monitor the vehicle traffic and sensors that check the occupation of the parking lots of Barra beach.
The various services implemented in this project can be accessed at this web portal. The portal allows a better perception of the monitored places, most of all, allows a better informed reflection about the places to visit or the roads to travel.
It is readily possible to understand two groups that will have great benefit with our project.
On the one hand, the transport infrastructure operators, this entities will be able to collect useful information about the state of the roads, wrong behavior on the part of drivers, as well as pollution statistics associated with an specific area.
On the other hand, ordinary citizens can also take advantage of our services in order to find out, for example, if it is worthwhile to go to a certain place, given that it may be affluent and therefore is overcrowded. Or something as simple as checking the traffic in a certain area.
In addition, if a citizen has a car equipped with technology that allows c2i-i2c communication, then she/he can also receive important information about the variables associated with the trip. Such as, climatic conditions or behavior of other road users that may put their own health at risk, as is the case of users who are speeding, therefore, requiring greater care.
To summarize, we present the following list with the main actors and a brief description:
-
User of the web application
-
This actor uses the web application to check the traffic in a certain place or just to verify the weather in a zone. The web application doesn't have authentication, so anyone who needs information about the traffic can just open the browser, type our url, and prove whatever was needed. So, both groups enunciated before belong in here.
-
User with the OBU (c2i-i2c)
-
This actor will have a technology in his car that will allow sending data directly to a dedicated infrastructure. After, the user will have access to data on what can be expecting during his trip. Weather conditions or dangerous behavior on the part of other road users are some examples.
To help the exposition of the use cases, 3 personas will be used:
-
João is a representative from Ascendi, a transport infrastructure operator from Portugal. His job is to check information related with the roads that Ascendi is responsible for. Having that in mind, João uses frequently the web application made available by us.
-
José is also a frequent user of the web application. In his case, he uses it because of his occupation. The transport of goods implies a lot of travel between locations, often needing to check road traffic, and which are the best roads to travel without major consternations.
-
Paulo is very fond of the beach, with the particularity of not really enjoying being close to large crowds. In this way, Paulo uses the web application in order to check which is the beach affluence, making the decision if it is worthwhile to go, according to the current capacity.
To better understand what the users can expect from our system, it will be presented below the different use cases associated with the distinct interactions that can be done using the web application.
Number | Use Case | Description | Priority |
---|---|---|---|
1 | check number of vehicles currently circulating | José goes to the web app. As the tab shown is the one of the dashboard, he just has to select which is the location he wants to check the traffic on, and at that moment he decides whether it makes sense or not to go that route. | High |
2 | see information about the traffic | In order to better understand if the roads monitored by João are in good condition or not, namely if they have been in a lot of traffic. João has the possibility to check this type of information from the ROAD INFO tab. | |
3 | check weather and adverse weather conditions | José is not sure if he can take an open box van to deliver a certain order. To be sure that there will be no surprises, he goes to the app, and visualize the updated climatic environment of the area. | High |
4 | being alerted when some driver is being careless | José needs to pay close attention to the variables on the road. With that in mind, José uses the ROAD DANGER EVENTS functionality to be alerted whenever a road user is being careless, traveling for example at excessive speed. | High |
5 | verify if there are animals on the road or other strange objects | If there are accidents on the highways that are caused by foreign objects on the road or by animals, the alleged culprits for that accident are the responsible entities for that road. Taking this into account, João uses the web application in order to be alerted when any type of foreign object is detected on the road. Acting quickly in accordance with what has been warned, avoiding greater evils. | Medium |
6 | check the influx to a given area | Paulo had a sudden urge to go for a run on the beach. But that will only be possible if the area is not overcrowded. So, Paulo opens the browser and goes to our web application. What he sees on the dashboard after selecting the desired zone is that the beach affluence is low, so he can go on a race without worries. | Medium |
7 | see carbon footprint | Environmental impact is a major concern these days. The transport infrastructure operators have a lot to gain by monitoring the carbon footprint of their highways. The sustainability stamp is a good indication of social responsibility. Following this thought, João uses the CONDITIONS tab of the web application to check the carbon footprint for a given highway. | Medium |
8 | check stopped cars on the road | Once again, José needs to move around constantly, it is something that is inherent in his profession. Therefore, he is constantly checking for risks on the roads he will be driving, such as accidents or cars stopped due to a breakdown. For this, he uses the functionality that alerts for road risks in the ROAD INFO tab. | Low |
Note: Concerning the second actor, User with the OBU (c2i-i2c). The referenced actor, doesn't have exactly an interaction with a system. What does he have, is the need to connect the OBU to the obd2 port of the vehicle, and after that, the sending of data to the infrastructure is automatic, so, no interaction needed.
With regard to business requirements we can enumerate the following:
-
An application that allows to take advantage of 5g technologies, as well of cameras and radars installed on Portuguese highways in order to monitor traffic and other situations that arise from it
-
Another technology with the intent of communication between cars, using the obd2 protocol and once again 5g intercommunication. It will be possible to pass information between cars, so that, risk situations are avoided or minimized. In addition to reducing risk situations, we also aim to make an estimate regarding vehicle pollution.
At this section we describe the services that we intend to provide to the community. Those services will all be delivered via web application. Below is a list we the requirements.
-
Provide a dashboard were the client can see information related to:
-
weather conditions by zone
-
traffic inflow
-
number of vehicles circulating a specific road, also associating data related to speed
-
number of cars caught speeding
-
some statistics associated with physical activities made by people on bike lanes, like walking or riding a bike
-
alerts about danger behaviors on the roads
-
allow the visualization of the carbon footprint for a specific zone
-
-
The use of 5g communications is mandatory, in order to reduce the latency between devices.
-
The management of the VNF’s associated with the cameras must be designed using a elastic approach, meaning that, VNF's should be created and erased as appropriate.
-
Radars must be used in addition to cameras, with the purpose of having sensor fusion, which will allow to obtain more reliable results.
-
It's necessary to have a web application that allows the display of the information collected through the various technologies
-
It should be followed a service-oriented development
-
The web application will be available for everyone, without the need to register and create login.
-
So, with regard to levels of authorization, it can be said that there is only one authorization level, that applies to all.
-
The web application must be easy to use
-
The user must be able to perform all possible tasks without the need for help
-
For the user to reach the different workflows he should not need to make many clicks
-
The web application will be open so there will be no need to login
-
With regard to data, it must be isolated on a machine
- The code must be robust and tested extensively in order to increase the reliability of the system, preventing applications from failing due to bugs
- The system must perform well as long as the continuous response and information update is maintained when dealing with continuous data streaming
-
The system must work 24/7, always providing all available services
-
As the implemented architecture follows a service orientation, if it is necessary to replace one of these services, this exchange should not significantly impact the normal behavior of the service in question
-
It is necessary to bear in mind that as traffic cameras are added to the system, it is also necessary to increase the computational power, namely the number of VNFs to analyze the streaming of the cameras
-
It is necessary to be careful with the workload that is being requested by the backend logic, if necessary, the backend must be replicated in different machines so that the computational power satisfies the system requirements
-
Documentation is very important for a good understanding of the project in general
-
API documentation becomes extremely important, given that it can be used by a client application without any affiliation with our project
To understand how our system will work as a whole, we defined the following technological architecture, that shows how the different layers and the components inside them are linking together.
The deployment diagram describes pieces of hardware to be used and the interactions between them. All the technologies are relatively standard, with the exception of the use of VNF to the deployment of the various modules.
The domain model tries to describe do conceptual model of the domain, that incorporates the behavior and data.