Abstract
The advancement toward 4G wireless technology (IMT-Advanced) introduced highly complex Geometry-based Stochastic Channel Models (GSCM), making the verification of simulation software a significant challenge. This paper proposes a rigorous vector-based verification framework that separates deterministic mathematical logic from stochastic random number generation. By utilizing standardized input vectors, the methodology allows for a direct comparison between a Software Under Test (SUT) and a reference implementation, ensuring precise adherence to international standards like ITU-R M.2135.
The Challenge of GSCM Verification
As wireless standards evolved, the models used to simulate radio environments became increasingly intricate. Older models were primarily defined by statistical characteristics, such as power delay profiles, which were relatively easy to verify by checking the output's second-order statistics.
In contrast, modern ITU-R M.2135 models are GSCMs that describe the generation of random waveforms through a series of complex, sequential functional blocks. Because these models involve numerous random variables, a subtle coding error might not be detectable by simply observing the final statistical output. There has been a critical need for a standard, rigorous way to prove that simulation software accurately reflects the underlying mathematical standard.
Vector-Based Verification Methodology
The core of the proposed solution is a vector-based software verification methodology based on the principle of separating "logic" from "luck."
The Separation Principle
The methodology decomposes the channel implementation into two distinct component types:
- Deterministic Functional Blocks: These represent the mathematical logic of the model.
- Independent Random Number Generator (RNG) Blocks: These handle the stochastic elements.
Verification Process
The verification follows a structured approach:
- Deterministic Testing: Instead of allowing the software to generate its own random numbers, researchers feed identical "input vectors" (pre-determined random seeds) into both the new implementation (the SUT) and a trusted reference implementation (such as the publicly available Matlab code).
- Vector Matching: If the deterministic logic is correct, the output vectors from both implementations must be identical for the same input. This allows for precise, block-by-block debugging.
- Statistical Validation: Once the individual blocks (SS-1 through SS-7, covering parameters like delays and angles) are verified, the Probability Density Functions (PDFs) of the final outputs are compared to ensure the overall statistical behavior matches the theoretical model.
Implementation and Results
The authors applied this methodology to verify a custom C-based simulator against a Matlab reference code. The evaluation focused on standardized deployment scenarios defined in the ITU-R M.2135 report, including Indoor Hotspot (InH), Urban Micro-cell (UMi), Urban Macro-cell (UMa), and Rural Macro-cell (RMa).
The verification was successful on two fronts:
- Quantitative Accuracy: For identical input vectors, the deterministic blocks in the C-based SUT produced output vectors that matched the reference exactly.
- Qualitative Alignment: Comparison of the Angle of Arrival (AoA) PDFs for Line-of-Sight (LOS) cases showed that the curves for the C-based SUT and the Matlab reference were virtually indistinguishable across various scenarios (InH, RMa, and UMa).
Significance and Future Directions
The primary insight of this work is that a stochastic system can only be verified with mathematical precision when the randomness is treated as a static input. This approach ensures that evaluations of competing technologies (like LTE-Advanced vs. WiMAX) are fair and not skewed by implementation errors in the simulators.
While the current work focuses on small-scale parameter generation, the principle is extensible to large-scale parameters and final Channel Impulse Response (ChIR) generation. As the industry moves toward 5G and 6G, which involve Massive MIMO and mmWave technologies, scaling this verification methodology to handle higher dimensionality will be a key area for future research.
Summary updated