GP-Driven Adaptive Tube MPC for Communication-Preserving Navigation of Mobile Relay Robots in Indoor Disaster...

Area of Science:

• Robotics
• Artificial Intelligence
• Wireless Communication

Background:

• Mobile relay robots face challenges in disaster environments due to signal degradation and narrow passages.
• Maintaining reliable communication and collision-free motion is critical for robot navigation.

Purpose of the Study:

• To propose a Gaussian Process-Driven Adaptive Tube Model Predictive Control (GP-ATMPC) framework for communication-preserving relay navigation.
• To address the disruption of multi-hop connectivity caused by non-line-of-sight (NLOS) degradation and structural bottlenecks.

Main Methods:

• Utilized Gaussian process regression (GPR) to create a probabilistic spatial radio map from RSSI measurements.
• Represented motion uncertainty using an adaptive ellipsoidal error tube.
• Tightened obstacle and communication constraints using a tube-tightened lower confidence bound (LCB) accounting for radio-prediction and motion-tracking uncertainty.

Main Results:

• The GP-ATMPC framework achieved the highest connectivity satisfaction rate among controllers preserving a safe motion margin.
• Demonstrated significantly fewer connectivity violations compared to nominal and heuristic MPC methods.
• Achieved millisecond-level online solve times and validated radio-prediction accuracy with a mean absolute error of 2.1 dB.

Conclusions:

• Coupling spatial radio prediction with adaptive tube-based robust control offers an effective solution for resilient communication-aware navigation.
• The framework provides a robust approach for mobile relay robots in degraded indoor disaster environments.
• The GP-ATMPC framework balances connectivity, safety, and computational efficiency.