1 of 8

Week 3 Progress Update

AI-Driven Autonomous Communication Discovery and Link Selection for Lunar Surface Rovers

Jonathan Morcos

WINLAB Summer Internship

Week 3

Results + modulation learning

16-QAM / OFDM

Focus: connect link-selection simulation with basic RF signal generation concepts.

2 of 8

Simulation Results

WINLAB weekly update

Selected link share

LTE/5G

85.30%

Earth

8.70%

Relay

3.80%

Orbiter

2.20%

91.60%

Availability ratio

8.40%

Outage probability

208.21 ms

Average selected delay

24.58 dB

Average selected SNR

How to read this slide

Left: how often each link was chosen. Right: overall quality of the selected links. The main result is that LTE dominates, but outages still happen when no link is usable.

3 of 8

Result Plots

WINLAB weekly update

Selected link over time

Shows which link the rover picked at each time step.

Delay over time

Spikes mean the selected link had higher delay.

SNR over time

Higher SNR means a cleaner selected signal.

Selection counts

Confirms LTE/5G was chosen most often.

Threshold vs. handovers

Shows handovers dropping as switching becomes stricter.

4 of 8

16-QAM Learning Demo

WINLAB weekly update

What I built

Generated random bits in Python.

Mapped groups of 4 bits into 16-QAM symbols.

Placed the symbols onto OFDM subcarriers.

Saved plots for constellation, waveform, and spectrogram.

16-QAM

16 possible points

Why this matters

The rover project is about choosing communication links. This RF demo helps me understand what digital communication signals look like once data is actually modulated onto a waveform.

5 of 8

16-QAM Constellation

WINLAB weekly update

How to read it

Each point is a possible symbol.

Horizontal direction is the In-Phase (I) component.

Vertical direction is the Quadrature (Q) component.

The receiver estimates which point was sent.

Key idea

A constellation is like a map of digital symbols. Cleaner clusters usually mean the receiver can decode the data more reliably.

6 of 8

OFDM Waveform

WINLAB weekly update

Real part of OFDM waveform

What happened in the code

Used 64 OFDM subcarriers.

Used an IFFT to convert frequency-domain symbols into a time waveform.

Added a cyclic prefix of 16 samples to each OFDM symbol.

What the waveform shows

The waveform is not supposed to look like individual bits. It is the combined signal after many subcarriers are added together in time.

7 of 8

RF Spectrogram

WINLAB weekly update

RF spectrogram of OFDM signal

What the plot means

x-axis: time, so we can see when the signal is present.

y-axis: frequency, so we can see where the signal energy is located.

color: power in dB, where brighter colors mean stronger energy.

This is closer to what SDR tools show: energy over frequency and time.

8 of 8

Week 3 Takeaway

WINLAB weekly update

Main takeaway

The link selector results show which communication option the rover chooses over time.

The 16-QAM constellation shows how bits become modulation symbols.

The OFDM waveform and spectrogram show how those symbols become a signal over time and frequency.

Next step: connect simulated link selection ideas with more realistic RF/SDR measurements.