Abstract:Wireless communications and sensing (WCS) establish the backbone of modern information exchange and environment perception. Typical applications range from mobile networks and the Internet of Things to radar and sensor grids. The incorporation of machine learning further expands WCS's boundaries, unlocking automated and high-quality data analytics, together with advisable and efficient decision-making. Despite transformative capabilities, wireless systems often face numerous uncertainties in design and operation, such as modeling errors due to incomplete physical knowledge, statistical errors arising from data scarcity, measurement errors caused by sensor imperfections, computational errors owing to resource limitation, and unpredictability of environmental evolution. Once ignored, these uncertainties can lead to severe outcomes, e.g., performance degradation, system untrustworthiness, inefficient resource utilization, and security vulnerabilities. As such, this article reviews mature and emerging architectural, computational, and operational countermeasures, encompassing uncertainty-aware designs of signals and systems (e.g., diversity, adaptivity, modularity), as well as uncertainty-aware modeling and computational frameworks (e.g., risk-informed optimization, robust signal processing, and trustworthy machine learning). Trade-offs to employ these methods, e.g., robustness vs optimality, are also highlighted.
What problem does this paper attempt to address?
This paper attempts to solve the uncertainty problems in Wireless Communications and Sensing (WCS). Specifically, the paper focuses on how to deal with uncertainties from multiple sources during the design and operation processes to ensure the reliability, resilience, adaptability and robustness of the system. The following are the main problems mentioned in the paper:
1. **Channel Uncertainty**:
- Channel modeling errors (e.g., mismatched modeling assumptions)
- Moving objects
- Environmental evolution and fluctuations
- Estimation errors due to scarce data and limited computational resources
2. **Noise, Interference and Attacks**:
- Thermal noise, quantization noise, phase noise
- Co - frequency interference, adjacent - frequency interference
- Inter - system interference (e.g., between Bluetooth and Zigbee)
- Self - interference (e.g., in full - duplex systems)
- Attack behaviors such as interference, eavesdropping, and deception
3. **Hardware Defects**:
- Power amplifier nonlinearity
- I/Q imbalance
- Phase noise
- Array calibration error
- Limited converter resolution
- Clock drift, etc.
4. **Deployment and Configuration Uncertainty**:
- Position errors of base stations and access points
- Deployment errors of sensor nodes
- Array calibration error
- Formation errors (e.g., in highly mobile UAV networks)
5. **Timing Errors**:
- Inaccurate synchronization between devices (e.g., between transmitters and receivers)
- Clock drift of oscillators
- Clock skew between different devices or systems
- Time quantization error
6. **Network Topology Changes**:
- Node movement, node availability (joining and leaving)
- Changes in network protocols
7. **Variations in Available Resources**:
- Optimization of the allocation of time, space, spectrum, power and computational resources
- Resource variations caused by environmental factors (e.g., weather, disasters)
8. **Environmental Uncertainty**:
- Rain, snow, humidity and temperature fluctuations
- Natural disasters, solar activities, space radiation
- Industrial electromagnetic interference, vehicle movement, human activities, etc.
9. **Algorithm and Computational Errors**:
- Model surrogate errors (e.g., sample average approximation of mathematical expectation)
- Local optimal solutions in optimization
- Empirical setting of initial and termination conditions of iterative algorithms
- Numerical errors (e.g., rounding, truncation, matrix inversion, quantization, discretization)
- Computational resource constraints (e.g., incomplete or delayed computation)
10. **Data Scarcity**:
- Inherent system limitations (e.g., sparse sensor deployment, limited sampling rate)
- Environmental limitations (e.g., harsh or dynamic environments)
- Privacy and security considerations (e.g., data restrictions)
- Data aging or high acquisition cost
Ignoring these uncertainties may lead to serious consequences, including performance degradation, untrustworthy systems, low resource utilization efficiency and security vulnerabilities. Therefore, the paper aims to enhance the system's awareness of uncertainty by reviewing mature and emerging architectures, computational and operational countermeasures, thereby ensuring the credibility of wireless communication and sensing systems.
### Solution Overview
The paper proposes two complementary strategies to deal with these uncertainties:
1. **Uncertainty - Aware Design of Signals and Systems**:
- Redundancy, diversity, margin
- Feedback and adaptation
- Anomaly detection and handling
- Modularity
- Distributed processing
- Prediction and prescription
2. **Modeling and Computational Frameworks for Uncertainty**:
- Uncertainty quantification
- Uncertainty - aware optimization
- Adaptive and robust signal processing
- Trustworthy machine learning
These methods aim to balance different characteristics of the system, such as adaptability and robustness, and carefully plan the trade - offs of these characteristics in practice.