Author | Jessie
Produced by | Yanzhi
5G promises to bring new innovations and applications to the automotive industry, but it is not just around the corner, and the automotive industry will have a connectivity "innovation gap" that is best filled with LTE-V – a flexible and dedicated solution for future vehicle communications.
In practice, the best-performing solution for C-V2x may be a communication system that combines sensors and cameras, supplemented by a high-definition map system that in turn receives real-time updates over a cellular network, as well as direct vehicle-to-vehicle communication with ad hoc network functions. (Due to mobility-related factors, such as travel speed and channel characteristics, the requirements for direct vehicle-to-vehicle communication vary from device to device.) If the hop count becomes large due to the protocol, the Ad hoc network will become very inefficient. One practical limit is the five-hop. If there is an active antenna system in front of and behind the car, the number of hops can be doubled. ）
The information provided by V2X technology is critical for future connected and autonomous vehicles to efficiently and safely navigate to their intended destinations. It's important to understand the amount of data that connected cars will transmit and receive, as this places very high demands on network capacity. Some estimates suggest that by 2020, each connected car will generate more than 4,000 GB of data per day. While connected cars are rapidly becoming mainstream, there are still questions such as where, what to do, and what connectivity is needed.
The application of autonomous driving scenarios of vehicle-road collaboration has been improved
1- The ability to automatically drive in basic scenarios is improved
The C-V2x-based vehicle-to-road coordination strategy improves the overall automatic driving function as shown in the following figure:
There are many ways to collaborate on autonomous driving development. A typical test case is automated lane consolidation to maximize road capacity. In order for overtaking vehicles to most effectively re-enter slower lanes in times of traffic congestion, it is ideal to have any vehicle in front of you accelerate slightly and slow down the vehicles behind to make enough space for the merged vehicles. The same process is desirable when vehicles enter a dense highway. For emergency trajectory planning, each vehicle broadcasts its identity, location, speed, and direction, and uses this data to build its own map and determine if any other vehicles are on a potential collision trajectory.
2 - 5G-based formation driving
Automatically connect trucks, vans, or cars to fleets that are safer than human drivers to save fuel and improve the efficiency of cargo transportation. It establishes a flexible arrangement on the highway and then decomposes as the vehicle leaves the highway. Direct communication between sensors and direct neighbors can be used to build a fleet of 2 or 3 vehicles. For longer horizontal queues, the spread of the message takes a long time. Braking must be synchronous and require low-latency network communication. For the orchestration of more than 3 cars, there will be higher requirements for 5G. The associated car needs to communicate with very low latency and acknowledge receipt of any messages.
In remote driving, it is mainly how to avoid obstacles in the remote state of the over-control car. Obstacles include lanes that have been blocked due to recent accidents, dual-vehicle parked cars that are not allowed to pass without crossing the yellow line at the entrance/exit, and ensuring that inexperienced vehicles are blindly advancing when safe action cannot be determined or do not know what is ahead. When the vehicle encounters such a situation, it stops or finds a location with minimal risk and then requests the assistance of a remotely controlled operator to control and bypass the obstacle. In order for the remote control to understand obstacles and determine the path the vehicle must take, the controller utilizes the streaming sensor information (e.g., video, lidar, radar) temporarily provided to him or her. Once the obstacle is cleared, the flow to the controller stops, and the performance requirements are:
The vehicle will have full control over its destination. The solution will require precision, and restrictions will need to be made to ensure high customer satisfaction and limit the traffic obstacles that will result from stopping autonomous driving.
The following are some of the key V2V communication requirements for supporting queues:
End-to-end communication latency between a group of vehicles of 25 ms (up to 10 ms automation)
90% message reliability and 99.99% maximum automation
The accuracy of the relative longitudinal position is less than 0.5 m
Broadcast rate of 10 to 30 messages per second
Dynamic communication range control to improve resource efficiency at different platoon sizes and limit message distribution for privacy reasons
Several aspects of formation must be supported by reliable V2V communication:
Joining and leaving the fleet: Allows vehicles to signal to join or leave the platoon at any time during platoon activity, and supports additional signals to complete the join/leave operation
Announcements and Warnings: Indicates the formation and presence of platoons so that nearby vehicles can choose to join the platoon or avoid interference with the platoon
Steady-state fleet group communication: support the exchange of pipe information, but also can indicate brakes, acceleration, which road to take, change the queue, etc.
Given the small target distance when the vehicle is traveling at relatively high speeds, V2V communication must be able to support reliable, high duty cycle and secure message exchange to ensure efficient and safe queue operation.
3- 5G-based remote driving
The vehicle is driven by a remote area, the vehicle is still driven by a person - there is no actual driver in the car. This may be used to provide premium concierge services, enable someone to attend meetings or work on the go, or support taxi services, or help people without a driver's license, or when they are sick, drunk, or otherwise unfit to drive. This process overall requires a highly reliable radio link with a full round-trip delay of less than 10 ms that is fast enough. So that the system can receive and execute instructions as quickly as the human eye perceives changes.
Performance requirements:
Allows human operators or cloud-based applications to remotely control vehicles via V2N communication. There are a variety of scenarios where remote driving can be leveraged, including:
Provide backup solutions for autonomous vehicles. For example, during the initial autonomous vehicle deployment, a remote operator can take control when the vehicle is in an unfamiliar environment and difficult to navigate.
Remote driving services are available for teenagers, seniors and others who do not have a driver's license or are unable to drive.
Enable fleet owners to remotely control their vehicles. Examples include moving trucks from one location to another, providing car hire services to customers, and providing remotely driven taxi services.
Enable cloud-driven public transportation and private shuttles, all of which are particularly suitable for services with predefined stops and routes.
Due to lower technical requirements (e.g., fewer on-board sensors and fewer computational requirements for complex algorithms), remote driving can be used to reduce the cost of fully autonomous driving for certain use cases. Here are some potential V2X requirements to support remote driving:
Up to 1 Mbps for downlinks and uplinks up to 25 Mbps (assuming up to 10 Mb/s each for two H.265/HEVC HD streams)
Ultra-high reliability (URLLC) of 99.999% or more
The end-to-end latency between the V2X application server and the vehicle is 5 milliseconds
Speeds of up to 250 km/h
When an obstacle blocks a Level 4 or Level 5 autonomous vehicle, remote control/driving will be required, making it impossible to decide on the path or method of safe navigation.
4- 5G-based environmental data processing
a. Perspective, Sensor Sharing/Camera Sharing
Sharing sensor data and camera images between vehicles enables the car to effectively "see through" other vehicles in front of it. A heads-up display (HUD) or augmented reality display on the driver's windshield combines what the driver can see with what the vehicle in front of it can see. This process requires that the HD video streams be perfectly synchronized and that time alignment is critical, so a very low-latency network is required.
Potential communication requirements between two V2X nodes that support extended sensors include:
High bandwidth, supporting burst transmission of large amounts of data
Latency is less than 10 milliseconds
95% high message reliability
High connection density to support congested areas (e.g., 15,000 cars per mile)
b. Environmental data processing - augmented reality mapping
Camera information will be overlaid on top of existing digital models of the surrounding environment to build a high-definition 3D map. Stereo cameras from multiple cars upload images to the cloud and overlay multiple images in a collaborative oversampling manner to create very sharp 3D landscape images. 3D images can also include infrared detail, an augmented reality mapping method that promises to produce better maps than existing services, and cars can compare 3D models to reality to recognize the differences between stored models and real-time images (identifying pedestrians, animals, cars, motorcycles, and details like changing surfaces on streets).
In this process, video needs to be transmitted from the car in real time and uninterrupted, and the centralized map data is received on the car in time for comparison. During the process, you will need a high-throughput network, such as enabling images for time synchronization time with as little time as possible.
Self-driving cars generate more than 4 TB34 of data every day through a variety of sensors such as radar, lidar, and cameras. These datasets are used for algorithms at various stages of vehicle system development and consumer development. During the development/pilot phase, most of the sensor data is stored in in-vehicle storage and transmitted to a data center platform to develop various deep learning models, which are then deployed in the vehicle for flexible detection and classification. Once the vehicle is sold, these models are periodically adjusted using new datasets from real-time driving. Depending on specific data patterns such as sensor data, vehicle diagnostics, positioning data, and real-time condition data, various wireless data uplink methods will be required.
c. Cloud-based application examples – telematics, mobile profiles, insurance and anti-theft
There are many ways to take advantage of vehicle mobility data (even in an anonymous form). For example, augment their traffic map by using data about the speed of movement reported by their devices. By analyzing places where the speed is significantly lower than the local speed limit, it can identify traffic congestion. Insurers can also collect mobile data and offer the option to install telematics equipment that reports driving patterns, speeds, and locations. Data is collected and analyzed in the cloud and used to determine the risks insured for each driver as well as the individual's premiums. For anti-theft, the location and driver of the vehicle can be tracked directly and this information is passed to the cloud for analysis. Where allowed, these applications can correlate data and determine where people work, live and shop in just one small step, and use that information to sell them the right services.
The network requirements for these applications vary from application to application to application, but in many cases, these use cases can be implemented with today's cellular networks because there are no very demanding latency or throughput requirements. Different types of communication are required between vehicles and other vehicles, transportation infrastructure, cloud-based and other applications, and even pedestrians or cyclists.
Performance requirements:
The potential communication requirements between two vehicles employing advanced driving communication are:
High bandwidth, support a large number of data burst transmission;
10 ms latency for maximum automation;
99.99% message reliability for the highest degree of automation;
d. Real-time HD mapping
Although technology implementation strategies vary, real-time, high-definition maps are a key element of autonomous driving. In the Map Lights method, high-definition maps are primarily used for navigation purposes and overlap with real-time situation data (e.g., accident notifications, road construction). In the "remap" approach, high-definition maps play a more critical role in path planning — even up to centimeter-level details. Such scenarios would require a map with a maximum size of 1 TB for a single city/community. Not only do these maps need to be updated regularly without direct user interaction, they may even need to be updated on demand as vehicles move across geographies. A deployment strategy that leverages roadside infrastructure and edge cloud solutions to deliver these map updates is critical to reducing costs.
C-V2x's principle of roadside communication
This article needs to focus on the communication principles and processes that are more typical in another V2x, and highlight how it can be applied to autonomous driving systems. These include typical way-end traffic movement principles.
The types of transmission data for each hardware device are as follows:
In order to clarify the cooperative relationship between V2X, OBU, RSU and V2V, we use the communication between vehicles and vehicles as an example to illustrate the overall working relationship and data flow transmission trend.
Suppose there are two intelligent driving cars A and B, each with hpc, an autonomous driving high-performance computing platform. The platform connects the various nodes of autonomous driving, including V2X nodes that forward data from OBU devices.
V2X device OBU. The roadside of the intelligent driving road is equipped with: road test unit RS (mainly used to send traffic lights, traffic signs and road obstacle data to the OBU on the body). Among them, the road test unit RSU and the V2X device OBU also communicate through radio waves. So how do multiple vehicles mainly communicate with vehicles and vehicles?
Suppose that the A car is the main perspective, and the data interaction with the B and C cars is viewed separately. Then the whole process is as follows:
1) Each ECU unit of A car subscribes to B and C car and roadside data
First, suppose that the v2x node on the HPC of a car A receives the Message sent by the OBU of B and C through the UDP protocol, which mainly includes the RSU data at the end of the road (mainly refers to the environmental data of the road end) and the information of the two vehicles B and C (including the body, chassis and positioning information, etc.). Then the v2x node on the A-car HPC parses the OBU message and packages it into an Autosar AP message and sends it out for all other nodes connected by the A-car HPC controller, in this process, it can be partially broadcast and partially subscribed to all nodes associated with HPC to do message subscription. The transmission of messages between nodes in the car is carried out in the previously mentioned SomeIP protocol format, of course, there is a process of encapsulating the UDP protocol (this process mainly requires SomeIP to join the packet header Header with its own protocol standard, and confirm whether it is the data wanted in development by 0x89 identification through the packet header similar to 0x88).
2) A car integrates self-driving data packets
Secondly, if the ECU associated with the A car HPC needs to transmit the chassis, body or positioning information from the car, it can directly package these AP messages into proto format, and then encapsulate the SomeIP protocol format in the form of UDP protocol, and then send the packaged UDP protocol package to the A car OBU through the v2x node (through the packet header similar to "obuxxx" and other identification to confirm whether it is the desired data).
3) Car A sends self-driving data to car B and C
A-car OBU sends the A-car's own information (body, chassis, positioning information, etc.) to the B and C OBU through radio waves, realizing the sharing of vehicle information of A-car B-car. The data flow of this process is sent by the two-way OBU, and after the obu of the car receives the obu data of other cars, it is directly packaged with the data from the RSU and transmitted to the v2x node of HPC.
Welcome to the Intelligent Transportation Technology Group!