6G-Enabled Distributed Intelligence for Humanoid Robotics: Joint Computing and Connectivity across Robot and Environment (Edge)

Topic description

Future humanoid robots will increasingly rely on computationally intensive functions such as perception, sensor fusion, environment understanding, planning, reasoning, and AI inference. While some of these functions must be executed locally to guarantee responsiveness, autonomy, and safety, the computational and energy constraints of robotic platforms may make it beneficial to execute other workloads at nearby edge infrastructure.

This PhD will investigate the interplay between in-robot computing, edge computing, and 6G connectivity for next-generation humanoid robotic systems.

The research will characterize representative robotic workloads and investigate when processing should remain inside the robot and when computation or data processing should be dynamically offloaded to edge resources. The decision will consider factors such as workload characteristics, available computing resources, hardware constraints, energy consumption, latency requirements, mobility, and the current state of the communication network.

Moving intelligence outside the robot, however, introduces a communication cost. The PhD will therefore investigate robot-to-environment communication as an integral part of distributed robotic intelligence. Different robotic workloads may impose substantially different requirements in terms of latency, reliability, throughput, determinism, and availability.

The research will explore how 5G-Advanced and 6G mechanisms, including differentiated QoS, network slicing, resource allocation, and potentially complementary IEEE-based wireless technologies, can support these requirements. Particular attention will be given to the dynamic interaction between application workload, available compute resources, and communication conditions.

The ultimate objective is to develop mechanisms for joint compute-connectivity decision making, allowing a robotic system to dynamically determine not only where a workload should execute, but also which communication resources are required to support that decision.

Key research questions

1. Which humanoid robotic workloads are suitable for local execution, edge execution, or dynamic partitioning between robot and edge?
2. How do workload characteristics, hardware constraints, energy consumption, and latency requirements influence computation placement?
3. What is the communication cost of moving sensor data, processing, or intelligence from the robot to the edge?
4. Which QoS characteristics are required by different robotic workloads in terms of latency, reliability, throughput, and determinism?
5. How can 5G-Advanced/6G network resources and network slices be dynamically adapted to the computing decisions of the robot?
6. When can complementary wireless technologies provide better or additional connectivity for particular robotic workloads?
7. How can computing and communication resources be jointly orchestrated to maintain robotic performance under changing workload, mobility, and network conditions?

Expected research outcome

The PhD will develop and experimentally validate a compute-connectivity framework for distributed robotic intelligence, enabling humanoid robots to dynamically exploit local and edge computing while accounting for the performance and resource cost of the underlying communication infrastructure.