5G teleoperated vehicles for future public transport
When autonomous mode fails, or human assistance is required in a complicated scenario like road blocked due to accident or work, the ability to monitor and remotely operate a driverless vehicle becomes crucial. In this post, we discuss Intelligent Transport Systems (ITS) and the significant roles cellular networks can play.
5G testbed to realize vehicle tele-operation
We are building a testbed for ITS in urban and controlled environments in collaboration with Scania, as part of activities within the Integrated Transport Research Lab (ITRL), and can already share results from tests with tele-operation of a bus. The results are taken from a bus at a test track and the feasibility of using a cellular network to realize vehicle tele-operation is assessed. Our testbed measurements reveal that tele-operation use cases can be supported even with less-than-excellent signal strength.
The testbed currently offers cellular connectivity for two sites: the Kista suburb of Stockholm and in Södertälje, Sweden. These two sites complement each other in terms of the type of ITS applications they will support: Kista is a semi-urban environment with people and vehicle traffic, which allows for realistic testing of ITS. Södertälje offers a safe environment for testing futuristic ITS applications such as remote driving.
The block component of the testbed in its current state.
Both radio access sites are controlled by an Evolved Packet Core (EPC) in Kista. EPC presently uses physical nodes, but we plan to make use of virtual network functions soon. An Open Stack-managed cloud in the same location hosts any off-board software the applications may use, in addition to a function that prioritizes data traffic of one ITS application over another. Data traffic prioritization is achieved by interfacing with the SAPC node, which is an implementation of the Policy and Charging Rules Function (PCRF).
Cloud computing: an enabler for 5G network
We developed a prototype system to address the challenge of operational complexity of 5G that uses cloud computing such as virtualization and network resource orchestration. Cloud computing is an enabler for 5G network operators to address above concerns and achieve economies of scale in ITS and IoT-at-large. Virtualization of ‘network resources’, such as nodes of the core and parts of the radio access network, reduces costs by enabling cellular network functions operation over commercial off-the-shelf infrastructure. Reduced time-to market due to automation in application lifecycle management and multi-tenancy (i.e. many applications using the same physical network resources) are also products of this virtualization. We introduced a lifecycle management function (LCM), which automates management of applications throughout their lifecycle – from deployment to operation and eventual decommissioning. We described this function as a concept in our paper presented at SOCNE 2015, Towards Automated Service-Oriented Lifecycle Management for 5G Networks, and included an initial implementation and early results in DevOps for IoT Applications using Cellular Networks and Cloud at FiCloud 2016.
Scania presented a demo on November 25 2016, in which they drove a bus remotely from a vehicle operations centre at Scania’s office in Södertälje using the testbed setup by Ericsson. You can see the demo in this video:
A remote operator drove the bus on the test track and back to the parking facilities, with speeds up to 20km/h. The test track is shown in the figure below. Sensor data on the bus, like the video feed from the forward-facing camera, is streamed to the remote station via the uplink of the cellular network.
The track of tele-operated bus in Södertälje.
Latency / data freshness measurements
We are currently in the process of evaluating the network connection of the test bed for use case deployment. As part of this activity, we measure latency and throughput between moving connected vehicles and the Ericsson cloud (see figure 1 for the setup).
Below, you can see an illustration of the results of these measurements performed at the Kista and Södertälje sites. For the greater part of the measurements, latency stays under 50 milliseconds, and there are only some network blind spots (e.g. sharp turns obstructed by natural or human-made obstacles such as hills or buildings respectively) that may increase the latency beyond this value.
measurements on the Kista (one-way-delay) (left) and Södertälje (round-trip-time) (right) test beds, between connected vehicle and Ericsson Cloud.
Latency of up to 50 milliseconds is sufficient for small-scale controlled tele-operated driving as well as mission parameter relay from off-board software to the bus. We currently use an LTE radio access network. However, we plan to upgrade our radio access network to 5G as technology becomes available, as well as install 5G modems on the Scania vehicles, which will further reduce the latency.
In addition to network latency, we are evaluating the data freshness of a video signal. Data freshness measurement includes the whole processing and transmission chain and is therefore a true end-to-end performance metric. For our application, we stream video from a camera inside the vehicle to a remote screen. Data freshness is defined as the time difference between i) an event happening in front of the camera, e.g., an obstacle appearing on the road, and ii) the event being visible at the remote screen. It includes: video capturing, video compression, data transmission over radio and transport networks, video decompression, and video rendering.
We measure data freshness by using two synchronized clocks with sub-millisecond resolution. One clock resides in the vehicle and is captured by the video camera. The other clock sits next to the remote screen. By taking a still picture of the screen, we can determine data freshness by subtracting the times displayed by the two clocks. Most measurements result in data freshness between 70 and 85 milliseconds. Also, the clear majority of measurement display a data freshness of less than 100 milliseconds. This is facilitated by using a set of hardware video encoder and decoder supplied by VITEC Video Innovations, which is optimized for low latency.
A proof-point of 5G for industries
A tele-operated bus from a command center presents a proof-point for the high expectations in 5G for industries. It is the first step in realizing as-a-service offerings for mobile network operators. Such offerings are expected to increase operators’ revenue and growth as connectivity is exposed to vertical markets that include but are not limited to automotive domain. Possible examples can be other ITS related use cases such as autonomous vehicles on public roads, ability to assist when autonomous mode fail on driverless vehicles, bus fleet management for passengers or truck fleet management to transport goods, cost savings by centralized remote-driver position. This will enable usage of cellular networks in new industries. Increased security from safe driver position (e.g. mining industry), increased productivity from ability to operate in dangerous environments (e.g. mining, under-water robots etc.), increased sensibility by human control with haptic feedback presents few examples.