# Space Ris Satellite Eval

> Evaluates a multi-agent DRL framework (MAPPO) combined with whale optimization for maximizing satellite coverage and data rates in 6G sub-THz networks using reconfigurable intelligent surfaces (RIS). Use when the user wants to benchmark on Simulated LEO Satellite-RIS Network Environment, or asks about evaluating this task. Reports average data rate.

- Skill: `qhjqhj00/space-ris-satellite-eval` (Agent Skill)
- Install (CLI): `npx skillmds add qhjqhj00/space-ris-satellite-eval`
- Raw SKILL.md: https://api.skillmd.com/api/skills/qhjqhj00/space-ris-satellite-eval/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: qhjqhj00 (https://skillmd.com/u/qhjqhj00)
- Updated: 2026-09-08
- Page: https://skillmd.com/skills/qhjqhj00/space-ris-satellite-eval

---


# space-ris-satellite-eval

> SpaceRIS: LEO Satellite Coverage Maximization in 6G Sub-THz Networks by MAPPO DRL and Whale Optimization — Sheikh Salman Hassan et al. (arXiv:2307.15469, 2023)

## What this evaluates

Evaluates a multi-agent DRL framework (MAPPO) combined with whale optimization for maximizing satellite coverage and data rates in 6G sub-THz networks using reconfigurable intelligent surfaces (RIS).

## Datasets

- **Simulated LEO Satellite-RIS Network Environment** — total ?; splits: test (-1)

## Metrics

- `average data rate` **(primary)** — range: other
  - Mean end-to-end data rate computed over all time slots and user equipment for a given distance or data packet size.
- `latency` — range: other
  - End-to-end data delay measured as a function of data packet size across different routing and phase-shift configurations.

## Input / output format

**Input**: State vectors containing 3D satellite coordinates, orbital parameters, RUE locations generated via HPPP, channel conditions, RIS element configurations, and current batch sizes.

**Output**: Action vectors for satellite-RUE association, data routing paths, RIS phase-shift angles, and GBS power allocation levels.

## Scoring recipe

```python
def compute_metrics(predictions, gold, env_state):
    total_data_rate = 0.0
    total_latency = 0.0
    n_samples = env_state['TE']  # 513 time slots
    for t in range(n_samples):
        rate, delay = env_state.simulate_step(predictions[t], gold[t])
        total_data_rate += rate
        total_latency += delay
    avg_data_rate = total_data_rate / n_samples
    avg_latency = total_latency / n_samples
    return {'average data rate': avg_data_rate, 'latency': avg_latency}
```

## Common pitfalls

- Assuming higher learning rewards directly translate to higher data rates without verifying convergence stability across distance ranges.
- Ignoring the non-linear threshold effect of RIS elements (N_r >= 20) where data rates only significantly improve beyond a specific element count.
- Misinterpreting the 'Central' baseline as a strong competitor; the paper notes it performs worst due to coordination overhead in dynamic environments.

## Evidence (verbatim from paper)

> Fig. 6 presents the important comparative analysis of the proposed MAPPO algorithm with several baselines... The evaluation involves assessing the average data rate as a function of the varying distance of data packets from the source satellite to the destination AoI. ... Fig. 7 presents results about one of the primary contributions of this research. The graph illustrates the latency (data delay) achieved by the proposed MAPPO algorithm concerning the data packet size, as compared to the performance of the baseline methods.

## Citation

```bibtex
@misc{hassan2023spaceris,
  title={SpaceRIS: LEO Satellite Coverage Maximization in 6G Sub-THz Networks by MAPPO DRL and Whale Optimization},
  author={Sheikh Salman Hassan et al.},
  year={2023},
  note={arXiv:2307.15469}
}
```

- arXiv: 2307.15469