Reliable Extreme Mobility Management for 4G, 5G and Beyond


Overview

(1) Problem:

We have witnessed a boom in various extreme mobility scenarios, such as the high-speed rails, vehicle-to-everything, drones, and many more. Compared traditional static and low-mobility scenarios, extreme mobility involve much faster client movement speed (up to 350km/h) in the outdoor environment. Despite this, many extreme mobility scenarios still need always-on Internet access “anywhere, anytime”. Today, a common solution is the mobile network, such as 4G, 5G and beyond. It is the largest wireless infrastructure that offers wide-area mobility management for network access. It has served billions of users today, and will hopefully serve trillions of mobile Internet-of-Things.

While the existing mobile network has been successful in supporting billions of mobile users, most users are moving slowly or static. With significantly faster client speed and 5G radios under higher frequency (e.g. sub-6GHz and above-20GHz millimeter waves), it is open to question whether existing mobility management design is still a good fit for extreme mobility.

(2) Objective:

This work explores whether existing 4G/5G remains reliable for extreme mobility, and we proposes a movement-based reliable extreme mobility management. We unveil various causes of unreliable 4G/5G in extreme mobility with real Chinese high-speed rail dataset from operational LTE network. We further propose REM, a movement-based reliability mobility management scheme for 4G, 5G and beyond.

(3) Findings:

We empirically study the unreilability of 4G/5G in extreme mobility. Our large-scale study in operational 4G LTE over high-speed rails unveils that, the mobility events are more frequent and vulnerable. On average, the handovers between base stations occur every 11–20s. Different from static or low mobility scenarios, handover failure and policy conflicts arise with alarming frequency: The network failure ratio ranges between 5.2% and 12.5% depending on the train speed, and the policy conflicts occur every 194–1090s. Both challenge the functionality of mobile networks and amplify the failures, delays, transient oscillations, and persistent loops. While the results are from 4G LTE, we believe 5G will face similar challenges with its same mobility management design as 4G LTE, adoption of millimeter waves, and denser small cell deployments with more frequent handovers.

We show that, the fundamental cause of unreliable 4G/5G in extreme mobility is its wireless signal strength-based design. 4G/5G mobility takes wireless signal strength as input, relies on the client- side feedback to trigger, and decides the target based on policies. While reasonable in static and low mobility, this design is sensitive to dramatic wireless dynamics from the Doppler shift in extreme mobility. Such dynamics propagate to all phases of mobility manage- ment and cause slow feedback in triggering, missed good candidate cells in decision, and unreliable signaling in execution. Our em- pirical study further shows that, operators have tried to mitigate failures with proactive policies. However, their methods amplify the policy conflicts and eventually offset their failure mitigation.

(4) Solutions:

We propose REM, Reliable Extreme Mobility management for 4G, 5G and beyond. Our key insight is that, client movement is more robust and predictable than wireless signal strength, thus suitable to drive mobility management. To this end, REM shifts to movement-based mobility management. The key enabler is REM’s signaling overlay in the delay-Doppler domain, which extracts client move- ment and multi-path profile with the recently proposed orthog- onal time-frequency space (OTFS) modulation. To relax the client-side feedback, REM devises a novel cross-based estimation by extending OTFS with singular value decomposition (SVD). It further simplifies the policy with provable conflict freedom, and stabilizes the signaling with a novel scheduling-based OTFS. REM is compatible with 4G/5G in static and low mobility, without changing their designs or data transfers.

(4) Outcome:

We prototype REM in commodity software-defined radio , and evaluate it with high-speed rails datasets and 4G/5G standard multi- path models. Compared to legacy solutions today, REM eliminates policy conflicts, reduces failures by up to an order of magnitude (0.9×–12.7× depending on client speed). Even in extreme mobility, REM achieves comparable failure ratios to static and low mobility scenarios. Meanwhile, REM retains marginal overhead of signaling traffic and latency without hurting data transfer.

Publication

Beyond 5G: Reliable Extreme Mobility Management  ACM SIGCOMM'20

Yuanjie Li, Qianru Li, Zhehui Zhang, Ghufran Baig, Lili Qiu, Songwu Lu
Annual conference of the ACM Special Interest Group on Data Communication (SIGCOMM)

PDF

Team Member

Artifact release

Download

Research Support

We gratefully acknowledge research support from the departmental support from UCLA and UT Austin.