Abstract:
Low-latency communication in 5G networks enables
advanced applications in networked robotics, such as
collaborative visual Simultaneous Localization And Mapping
(vSLAM), where robots share real-time sensor data to build
accurate maps and localize themselves. To reduce onboard energy
consumption, robots can offload compute-intensive tasks—such
as feature extraction and map fusion—to nearby Multi-access
Edge Computing (MEC) servers. However, existing 5G traffic
steering mechanisms often ignore the dynamic compute load at
MEC sites, leading to service congestion and degraded Quality
of Service (QoS). This paper proposes a novel traffic steering
scheme that jointly considers real-time network conditions and
MEC compute availability to optimize service instance selection.
Leveraging the 3GPP-defined Application Function (AF), our system
dynamically associates mobile clients with the most suitable
service endpoint, minimizing end-to-end latency and improving
load distribution. We outline an experimental evaluation plan
including key performance metrics, a 5G-enabled testbed, and
representative robotic workloads. The proposed scheme is based
on Release 18 and adaptable to future beyond-5G architectures.
