4 scoped briefs

2 mini · 2 major

Computer Networks & Protocol Simulation

In networking, the result IS the measurement — which is why student projects here fail so predictably: one simulation run, a NetAnim screenshot, and a bar chart with no baseline and no error bars, which collapses the moment an examiner asks what a different seed would do. We build the opposite — parameter sweeps, a stated baseline, and confidence intervals on every curve, with the simulation environment pinned so the figures regenerate. Any attack or failure modelling stays inside the simulator, on topologies we generate ourselves, and is never pointed at a live network.

ns-3OMNeT++ / VeinsSUMOMininetONOS / os-kenOpenFlowWiresharkiperf3Python + matplotlib

Start a project Discuss a custom brief

Two audiences, one engineering standard

Academic depth meets production discipline.

For students

Pick a brief below, or bring the problem statement your department handed you. We architect it with you, review every commit, deploy it to a real URL, and rehearse the viva until no question surprises you.

See student support

For companies

Protocol evaluation and capacity what-ifs before anything touches production hardware, SDN controller applications, V2X and vehicular network modelling, and turning an internal performance claim into a measurement someone else can reproduce.

Discuss a company build

Mini projects

Focused scope. Real engineering.

Three to four weeks. Narrow enough to finish, deep enough that you learn the decision that actually matters.

2 briefs

Mini project3–4 weeks

MANET Routing Under Mobility: AODV vs OLSR

A controlled ns-3 sweep of AODV against OLSR across node density, mobility speed and offered load, reported with confidence intervals instead of the one lucky run.

ns-3C++PythonmatplotlibWireshark

You walk away with

Sweep: 3 node densities x 4 mobility speeds x 3 offered loads, 20 seeds per cell
Six metrics per configuration — PDR, throughput, end-to-end delay, jitter, routing overhead, energy — each plotted with 95% confidence intervals
FlowMonitor output cross-checked against the pcap traces in Wireshark, so every headline number has two independent sources
Trend comparison against published AODV/OLSR studies — direction and slope only (does PDR fall with speed at the same rate?), with an explicit statement that absolute PDR and delay figures are not comparable across different PHY, propagation, mobility and traffic setups
Pinned config — ns-3 version, propagation model, mobility model, seed list — in one script that regenerates every figure

Mini project3–4 weeks

SDN Load Balancer as an OpenFlow Controller App

A controller app on Mininet that spreads flows across a server pool and reroutes around a failed link, with failover time measured from packet timestamps rather than asserted.

Mininetos-kenOpenFlow 1.3Open vSwitchiperf3

You walk away with

Controller app with two policies — round-robin and load-aware flow assignment — built on os-ken (the maintained Ryu fork); Ryu itself is documented as the legacy path, not used as the default, because it is frozen upstream and its eventlet dependency fights modern Python
Measured flow distribution across backends under iperf3 load, quantifying the imbalance produced by naive hashing
Link-failure injection with failover time measured from packet timestamps, not from controller log lines
Flow-table dumps (ovs-ofctl) proving rules are installed in the switch, not merely logged by the controller
Dockerfile pinning the Python and controller versions, so week one goes into controller logic and not environment archaeology

Major projects

Capstones you can defend.

Ten to twelve weeks. Architecture, trade-offs, failure modes, deployment, and evidence—the project that carries an interview.

2 briefs

Major project10–12 weeks

Reinforcement-Learning Routing Protocol

A routing policy that learns from link state and mobility, benchmarked against AODV on topologies it never trained on, written up as an IEEE-format paper with a harness anyone can re-run.

ns-3ns3-aiPythonPyTorchmatplotlib

You walk away with

Pinned toolchain fixed in week 1 — a specific ns-3 release paired with the exact ns3-ai version that supports it; the bridge is version-locked and a mismatch discovered in week 3 costs weeks
Tabular Q-routing running end-to-end through the ns3-ai loop by week 3, before any neural policy — this de-risks the bridge and doubles as a learned baseline
The policy implemented as a real ns-3 routing module, not a script post-processing trace files after the fact, with the state, action and reward specification written down
Stated and met training budget: episodes x sim-seconds, runs parallelised across seeds, wall-clock cost reported — RL over a discrete-event simulator is sample-hungry and the schedule says so up front
Benchmark vs AODV on PDR, delay, routing overhead and energy including held-out topologies; IEEE-format paper naming the scenarios where the learned policy loses to AODV rather than burying them

Major project10–12 weeks

VANET Safety-Message Dissemination on a Real City Map

Emergency-braking warnings broadcast between vehicles driving a real OpenStreetMap road network, coupled closed-loop over TraCI so a car that receives a warning actually brakes — studied on 802.11p, with the post-2020 C-V2X shift stated rather than ignored.

OMNeT++VeinsSUMOIEEE 802.11pTraCIPython

You walk away with

Real road network imported from OpenStreetMap into SUMO with OSM-routed demand — the flow assumptions (randomTrips parameters or an assumed OD matrix) are documented in the report and are not called calibrated, because there are no ground-truth counts to calibrate against
Closed-loop coupling: SUMO and OMNeT++ step together over TraCI under Veins, so a received warning changes the receiving vehicle's behaviour and reaction-time and near-miss metrics are legitimate — an open-loop mobility-trace replay could not support any safety claim at all
Obstacle shadowing wired in from the same OSM extract (building polygons via polyconvert into the Veins obstacle model), with the LOS/NLOS split reported alongside every distance-bucketed latency number — a plain log-distance model reports clean packets at 500 m through a building block, which is fiction
Contention-based broadcast suppression measured against naive flooding: delivery ratio, warning latency at 100/200/500 m, redundancy ratio and channel busy time, all against vehicle density
Rush-hour and off-peak scenarios, 15 seeds each, confidence intervals on every curve, plus a report section positioning 802.11p as the studied baseline against the C-V2X / NR-V2X deployment direction the 5.9 GHz reallocation set

Explore more

Other engineering domains

Search every brief

Python Full-Stack Java & Enterprise AI & LLM Applications Machine Learning Deep Learning Data Engineering Mobile Apps Cloud & DevOps Cybersecurity IoT & Embedded JavaScript Full-Stack VLSI & Chip Design Power Electronics, EV & Energy Systems Signal Processing & Wireless Communications Robotics, Drones & Autonomous Systems Data Analytics & Business Intelligence Test Automation & SDET

Your next build starts here

Turn the idea into software people trust.

Bring us a product brief, a business problem, or a final-year project. We’ll turn it into a clear scope, a working build, and a handover you fully own.

Start Your Project Send your brief

Prefer email? info@tenzok.in

Computer Networks & Protocol Simulation

ns-3OMNeT++ / VeinsSUMOMininetONOS / os-kenOpenFlowWiresharkiperf3Python + matplotlib