IMDA 5G use-case findings
As cities become more compact, people need to find new and more efficient ways to get from one point to another. Whether it’s to deliver goods or for personal mobility, common modes of transportation today still require humans to be in control of the vehicle. Vehicle manufacturers and solution providers are now racing towards the opposite end of the scale where human control of the vehicle is no longer required. For this capability, advanced connectivity with onboard vehicle intelligence becomes fundamental to the success of this vision. Bringing us one step closer to this, are fully autonomous vehicles1 (AVs), or driverless vehicles, where anyone and everyone can get to their destination safely without the need of human intervention.
This is the thought behind the 5G trial project with IMDA undertaken by CapitaLand2 in partnership with NavInfo3, which explores the autonomous vehicle space that’s paving the future of mobility. The concept was simple: to experiment and explore the possibilities that 5G can bring for AVs through cellular vehicle-to-everything (C-V2X4). C-V2X is the foundation for vehicles to communicate with each other and everything around them, providing 360-degree non-line-of-sight awareness and a higher level of predictability for enhanced road safety and autonomous driving.
“We want to drive the creation of a new mobility industry, one that does not harm the environment. At the same time, we want to address common first and last mile connectivity issues that we’re seeing in the logistics industries today,” says Aylwin Tan, Chief Customer Solutions Officer at CapitaLand.
Designing a 5G-enabled smart estate
The trial to test AVs with C-V2X, was conducted in Singapore’s largest 5G smart estate trial site, at the CapitaLand-owned Singapore Science Park5. It was also CapitaLand’s first 5G trial in Science Park, and one that would pave the way for future projects around the Internet of Things (IoT), smart buildings6 and eventually smart vehicles. It is hoped that the enhanced connectivity of 5G would be able to develop Science Park into a 5G-enabled Smart Estate, opening doors for many more trials of a similar nature.
“As a real-estate company, it is in our interest to explore how our infrastructures can support and complement the needs of AVs,” says Tan. He adds that gaining more knowledge about 5G and its performance requirements, would help the company support future use cases, including commercialisation projects such as food delivery or automated emergency services. “With electric AVs, we would also be able to encourage greener footprints for people who visit our property, be it for leisure or to go about their day-to-day activities,” he adds.
Running the 5G AV trial at the Science Park was an ideal move, as it ensured that tests could be done in an environment with parameters that were as close as possible to the real environment, for the best outcomes. This is because unlike the Pasir Panjang roads7 that are accessible to the public, the Science Park was relatively free from AV testing restrictions.
On public roads, a driver is required in the vehicle as an additional precaution, which defeats the purpose of a fully autonomous vehicle trial. However, being able to use the Science Park as our trial site gave us the autonomy to really put C-V2X vehicles to the test in a safe space.
Supporting AVs from the cloud with 5G
As a market leader in autonomous vehicle navigation and smart mobility solutions, NavInfo’s partnership in this trial aimed to discover more about designing 5G-enabled C-V2X solutions that can relay real-time critical information through a cloud-based infrastructure8. The mission was to test the feasibility of a project like this, and use those learnings to explore uncharted territories in the smart mobility sector.
The team designed a 5G-enabled C-V2X infrastructure across two sites of the Science Park facility, deploying 3.5GHz band standalone units at each site to provide the high capacity 5G coverage required.
The trial demonstrated that the 5G network could indeed handle real-time streaming of sensor data from vehicle to cloud, a finding that was crucial to further AV innovation. The latency in terms of mean round-trip time (RTT) ping9, was measured at 9.61ms on 5G as compared to 28.41ms on 4G. A degraded mean RTT of 12 to 14ms was observed over 5G, when the network was loaded with data ranging from 25Mbps to 90Mbps. During this test, the mean throughput on 5G was measured at 66.85Mbps for downlink, and 66.82Mbps for uplink, with a notable dip in performance of 68% - 41%, during handover of 5G cells. “One of the purposes of this trial was to validate the possibility of cloud integrated sharing of information. By shifting the processing capabilities to the cloud, we would be able to reduce the load on a single AVs, in turn reducing development cost for better commercial viability,” explains Lam.
Due to the limitations of an ecosystem that is not yet mature, the trial validated the initial part of this cloud integration process, which is the uploading of data, but could not confirm the ability to fully process the large amount of visual data collected by light detection and ranging (LiDAR10) and camera sensors, of the AVs. However, there was still a good amount of knowledge gained from this process, as it allowed the team to understand the type of sensors and resolution, which 5G networks could readily support.
The second part of this trial highlighted the feasibility of segmenting data-collection like obstacle perception and real-time map updates, between multiple sensor platforms. “Based on the data we’ve gathered, we would be able to optimise the amount of data processing to be carried out by each AV, so that the gathering of information can be split across multiple vehicles within the same vicinity.” says Lam. During our test, we noticed 5G could support real-time perception processing for at least 99% of the packets11 with data uplinks less than 60Mbps from a single AV. However, at 95Mbps for a single AV, there was significant degradation, where approximately 30% of data packets were dropped due to high latency from an overloaded network. To improve this, a fleet of AVs operating in the same area could mitigate the probability of such losses.
By optimising the amount of processing carried out, the bandwidth consumed by each vehicle is significantly lowered. The surrounding visual, would then be consolidated in the cloud, enabling a 360-degree view constructed from inputs of multiple AVs. This is beneficial both from an operational and commercial point of view, as the workload can be distributed across multiple AVs. Overall, the positive outcomes and possibilities of distributed cloud processing, has given the team confidence to further develop this capability.
Robotics essential for automation of the future
Having seen the benefits of 5G as a viable alternative to wired infrastructure, CapitaLand already has plans to equip their buildings with 5G networks for both enterprise and consumer connectivity. This would provide a seamless experience for customers looking for solutions transitioning both outdoor and indoor mobile use cases in the future.
“To apply learnings from AV, we’re also looking into robotic technology, where robots can support almost everything, from ensuring safe and efficient lift operation, to responding to everyday needs like knowing when and how to clean a spilt cup of coffee,” explains Tan. After all AV’s are a version of mobile robots, with transport capability. To enable this vision, the group is working to strengthen its infrastructure across several key facilities which it operates. “This is key before innovative robotic automation can happen, as this will enable a seamless end to end user experience without human intervention,” he adds.
Essentially, it’s about creating an ecosystem that can be fully automated. Agreeing with Tan, Lam says, “Certainly, this is also our vision and with trials like this, we pick up new learnings and take new steps forward to making this vision a reality.”
Footnotes
1 Autonomous vehicles - A self-driving vehicle, also known as an autonomous vehicle, driver-less or robotic vehicle incorporating vehicular automation, that is, a ground vehicle that is capable of sensing its environment and moving safely with
little or no human input.
2 CapitaLand is a Singaporean headquartered company focusing on investment, development and management of real estate. It is one of Asia's largest real estate companies and the owner and manager of a global portfolio comprising integrated
developments, shopping malls, lodging, offices, homes, business parks, industrial and logistics assets, as well as real estate investment trusts (REITs) and funds.
3 NavInfo is a technology company which develops products in map navigation, dynamic traffic information, navigation software, telematics, and autonomous driving.
4 C-V2X – Cellular vehicle to everything was developed within the 3rd Generation Partnership Project (3GPP) to replace DSRC in the US and C-ITS in Europe. Cellular V2X is a 3GPP standard for vehicle to everything applications, such as self-driving
cars. It uses 3GPP standardised mobile cellular connectivity, to exchange messages between vehicles, pedestrians, and wayside traffic control devices such as traffic signals.
5 Singapore Science Park - is a research, development and technologies hub in Queenstown, Singapore. Managed by Ascendas, a subsidiary of CapitaLand.
6 Smart buildings - converge various building-wide systems such as HVAC, lighting, alarms, and security into a single IT managed network infrastructure for intelligent automation.
7 Pasir Panjang Road – One of the main public roads in Singapore.
8 Cloud based infrastructure - is the on-demand availability of computer system resources, especially data storage (cloud storage) and computing power, without direct active management by the user.
9 Ping operates by means of Internet Control Message Protocol (ICMP) packets. Pinging involves sending an ICMP echo request to the target host and waiting for an ICMP echo reply.
10 LiDAR - Lidar is an acronym of "light detection and ranging” or "laser imaging, detection, and ranging". It is sometimes called 3-D laser scanning, as it targets an object or a surface with a laser and measuring the time for the reflected light to return to the receiver.
11 A packet consists of control information and user data, the latter is also known as the payload. Control information provides data for delivering the payload (e.g., source and destination network addresses, error detection codes, or sequencing information). Typically, control information is found in packet headers and trailers.
.webp)