PROBLEM STATEMENTS
Controls
1. Model a simple DC motor and control its speed using a proportional (P) controller. What issue
do you see here? Explain.a
(Hint: You can use Python/MATLAB for this and read sensor readings in fixed discrete steps and
plot it.)
2. Mid-flight, the battery voltage suddenly drops by 20%. The drone loses some thrust. What
sensor detects it ? How can you make sure it stays stable and doesn’t crash immediately?
3. Simulate a simple balancing system in [Link] with a variable angle that changes
randomly. Write a short code that applies a “correction” (e.g., subtracts part of the error each
step).Plot how the “angle” becomes stable over time.
4. Imagine a delivery drone flying in a crosswind. Design an algorithm that automatically adjusts
motor thrust to maintain straight flight.
5. A microcontroller reads multiple sensors in a loop. Sometimes it misses readings from one
[Link] would you structure the firmware logic to make sure all sensors are read
consistently?
Resources
[Link]
NluWc16
[Link]
Sensors and actuation
[Link] how PWM (Pulse Width Modulation) is used to control actuators.
Tasks:
1. Write code in Arduino/ESP32 (or simulate in MATLAB Simulink).
2. Control LED brightness using PWM with varying duty cycles.
3. Optionally plot brightness vs duty cycle in MATLAB.
2. Imagine the VTOL starts to tilt left.
1. Which sensor detects it?
2. Which actuator should respond?
3. Given a list of drone components (MCU, ESC, motors, sensors), research or assume their
power usage.
1. Make a power budget table (Voltage, Current, Power).
2. Calculate total current draw and battery capacity needed for a 10-minute flight.
4. Imagine you are routing traces on a 2-layer PCB. You have three signals: Power, Ground,
and Data.
1. Which should have the shortest path?
2. Which should be thicker?
[Link] Would You Make an LED “Smart”?
An LED normally just turns ON or OFF.
Imagine you want it to “respond” intelligently, for example, blink faster when the battery is low,
or pulse softly to indicate sleep mode.
1. Think of what inputs or signals it could use.
2. Decide how it can communicate status without text or sound.
3. Make a simple logic flow (no code).
GUIDANCE & NAVIGATION
[Link] and explain the purpose of Coordinate Frames (like Body Frame, Navigation
Frame, Earth-Centered Earth-Fixed (ECEF) frame) in drone navigation. Why can't a drone's
navigation system just use one single coordinate frame for all its calculations and sensor
readings?
[Link], in simple terms, what GPS RTK (Real-Time Kinematic) is and why it's considered
more accurate than standard GPS for drone navigation. How does it achieve this higher
accuracy?
[Link] Scenario: The G&N system needs to tell the drone which way to "point" to reach its next
waypoint. For simplicity, we'll calculate the 2D heading angle.
● The Task:
1. The drone is currently at Current Position: (North: 10m, East: 20m).
2. The next Target Waypoint is (North: 30m, East: 40m).
3. Calculate the North difference () and East difference () between the target and
current position.
4. Use the math.atan2(y, x) function (where y is and x is ) to find the heading angle
in radians.
5. Convert the angle from radians to degrees (1 radian = ~57.296 degrees) and
ensure the result is between 0 and 360 degrees. (North = 0/360, East = 90,
South = 180, West = 270).
6. Print the heading angle in degrees, rounded to two decimal places.
● Learning Resources:
○ Python math.atan2: Python math.atan2() documentation
○ Radians to Degrees: Online converter or basic formula
Heading Angle (Conceptual): Understand what a "heading" or "bearing" means. Wikipedia -
Bearing (navigation)
[Link] a short, critical moment where the drone flies directly under a metal roof, completely
losing both the GPS and NavIC signals simultaneously. The drone's G&N system must continue
generating the accurate State Vector for the Controls department.
If the only sensor remaining is the IMU (Accelerometers and Gyroscopes), will the resulting
position estimation error grow faster in velocity or in position? Explain your reasoning using the
fundamental principles of integration (no formulas needed, just the idea).
{for better understanding refer: IMU Fundamentals}
[Link] Scenario: Our drone's IMU reports velocity in its own Body Frame (e.g., how fast it's
moving forward relative to itself). For G&N, we need this velocity in the North-East-Down (NED)
Navigation Frame. A Rotation Matrix performs this job.
● The Task:
1. The drone's velocity in its Body Frame is meters/second (forward, right, up).
2. The rotation matrix (Body to NED) represents the drone's current orientation:
[Link] a Python function transform_velocity(body_velocity, rotation_matrix) that uses
NumPy to perform the matrix-vector multiplication:
[Link] the resulting vector, rounded to three decimal places.
Learning Resources:
● NumPy for Matrix Multiplication: NumPy dot() function for vector-matrix multiplication or
the @ operator.
● Understanding Rotation Matrices (Conceptual): Focus on what they do (change
orientation). Wikipedia - Rotation Matrix (2D/3D overview) (Just read the introduction and
see the 3D examples).
5G COMMUNICATION AND NETWORKS
Use matlab to solve the simulation related problems
🔗 Reference Paper
Problem 1: 5G Network Slicing for Control and Video Flows
Simulate dynamic resource slicing between Control & Command (C2) URLLC-like traffic and
video eMBB traffic to evaluate latency and reliability trade-offs.
Tasks:
1. Assume total bandwidth B = 20 MHz.
2. Implement two logical slices:
○ C2: 10–200 byte packets, 50 Hz frequency, latency target < 50 ms.
○ Video: variable bitrate (2–20 Mbps).
3. Model per-slice FIFO queue.
4. Compute latency distribution (CDF) and packet success ratio (reliability metric).
Measurements:
1. Latency for C2 vs. video slice.
2. Missed-deadline ratio (number of packets exceeding latency bound).
3. Impact of changing slice bandwidth share dynamically.
Problem 2: Energy Efficiency of UAV 5G Communication
Estimate the energy cost of 5G connectivity on UAV battery life during urban missions.
Tasks:
1. Assume baseline UAV power Pbase = 90 W; 5G transceiver adds P5G = 1.5 W.
2. Compute flight time using
3. Compare with baseline (no 5G load) and visualize the relative endurance reduction.
Outputs:
● Flight time vs. 5G Tx duty cycle chart.
Problem 3: Handover Performance in UAV Mobility
Simulate a UAV crossing between two base stations and evaluate handover latency and packet
loss under LoS and NLoS transitions.
Tasks:
1. Model two base stations with overlapping coverage.
2. Define handover trigger (RSS or SINR < threshold).
3. Simulate UAV movement trajectory; packets sent continuously during the handover.
4. Compare interruption time between LoS and NLoS zones.
Measurements:
1. Handover interruption duration.
2. Packet loss rate during transition.
3. Correlate with latency distributions.
Problem 4:
Create a report describing how 5G communication architecture can maintain reliable
Control-and-Command (C2) and high-bandwidth video data flows in dense urban airspace
characterized by multipath, blockage, and mobility.
Expected Discussion Points:
1. Analyze 5G network components (gNB, core, edge computing) relevant to UAV
integration.
2. Explain architecture modifications for UAV communication (3D cell planning, adaptive
beamforming, and handover optimization).
3. Discuss challenges such as Doppler shift, interference management in vertical coverage,
and achievable reliability for URLLC in flight.
4. Relate to hybrid network designs (5G + satellite + direct air-to-air links).
Problem 5:
🔗 Reference Paper
I:
1. Explain the core concept of 5G-enabled UAV communication focusing on how
low-latency control (C2) and high-throughput video streaming coexist through network
slicing.
2. Present and analyze simulation outputs such as latency, throughput, coverage, and
energy consumption, discussing their impact on UAV operations in urban environments.
3. Summarize the overall UAV communication architecture, highlighting important aspects
like mobility management, handover strategies, security, and QoS prioritization, and
explain how 5G supports reliable and efficient UAV control and data transmission.
4. Why are the evaluation metrics for the experimental study based on SS-RSRP and not
on SS-RSRQ and SS-SINR? What does this tell us about the setup? How is it relevant
from the point of view of the study?
AI/ML
Problem 1.1: Vehicle Routing Optimization with ML
Use machine learning to predict optimal delivery routes for multiple vehicles considering
traffic patterns, delivery time windows, and vehicle capacity constraints.
Dataset
Vehicle Routing Problem Dataset – Solomon instances
Deliverables
1. Route optimization algorithm implementation
2. ML travel time predictor with feature importance analysis
3. Performance comparison report (ML vs. traditional methods)
Problem 1.2: Shortest Path with Dynamic Weights
Implement and compare A*, Dijkstra’s, and a neural network approach to find shortest
paths in graphs where edge weights change over time based on learned patterns.
Dataset
NYC Taxi Trip Data for traffic pattern learning
Deliverables
1. Multi-algorithm path finder with modular design
2. Weight prediction model with temporal analysis
3. Performance benchmarks across different scenarios
Problem 1.3: Maze Navigation with Reinforcement Learning
Train multiple RL agents (Q-learning, SARSA, DQN) to navigate complex mazes and
compare their learning efficiency and final performance.
Dataset
DeepMind Lab Maze Dataset or generated mazes
Deliverables
1. Three RL implementations (Q-learning, SARSA, DQN)
2. Learning curve analysis with convergence metrics
3. Policy visualizations and behavior comparison
Problem 1.4: Clustering-Based Path Planning
Use unsupervised learning to identify optimal waypoints in 2D space, then connect them
using graph algorithms for efficient path planning.
Dataset
OpenStreetMap GPS Traces or Geolife Trajectory Dataset
Deliverables
1. Clustering pipeline with multiple algorithms
2. Graph-based planner with visualization
3. Waypoint optimization analysis and cluster quality metrics