I Believe I Can Land#

My First Engineering Project (CanSat 2019)#

In 2019, just as I entered high school, I joined a small team that decided to take on the CanSat competition-an international challenge where students design, build, and launch a miniature “satellite” the size of a soda can. Even though CanSats don’t go to space, they’re deployed from ~1 km altitude and must perform real scientific and engineering missions: gather measurements, transmit telemetry, survive landing, and accomplish a self-designed secondary mission.

This became one of my earliest-and most defining-technical projects. It combined software, electronics, mechanics, physics, and a huge amount of trial and error. Our team called the project “I Believe I Can Land.”

Team & Roles#

We were a small group of four students, supervised by our IT teacher.

Although we initially divided responsibilities into clear roles, in practice everyone ended up doing everything-coding, testing, wiring, designing, and fixing whatever broke (which was often). It was true hands-on teamwork.

• Lead Software Engineer - Krzysztof Zbudniewek
  Built the full software stack: firmware, Raspberry Pi software, and the ground station.

• Lead Electrical Engineer - Bartłomiej Jacak
  Designed the electrical system, wiring, and the integration of all modules inside the tiny CanSat cylinder.

• Lead Recovery System Engineer - Bartłomiej Krawczyk (me)
  Responsible for modeling, testing, and designing the parachute system, including attempts at a guided parafoil.

• Software Engineer - Mateusz Kwiatkowski
  Involved in developing firmware for the payload.

• Team Supervisor - Anna Stopińska
  Oversaw the project and helped us navigate the competition requirements.

We were a small team-but motivated, curious, and very inexperienced. This project became our crash course in real engineering.

Mission Objectives#

Every CanSat must complete a primary mission and a self-designed secondary mission. Our secondary mission was ambitious-far more ambitious than we realized at the time.

1. Primary Mission: Telemetry & Environment Data#

The CanSat needed to measure:

• temperature (DS18B20 sensor)
• static pressure (BMP280 sensor)
• flight altitude
• frame number / timestamps

Data had to be sent live to the ground station via an ATSAMD21G18-based microcontroller and stored on an SD card for post-flight analysis.

2. Secondary Mission: Real-Time Landing Site Classification#

Our bold plan was to classify terrain beneath the CanSat during descent to detect areas that were:

• safe (fields, grass, open terrain)
• unsafe (water, buildings, roads, forests)

We used:

• a Raspberry Pi Zero
• Pi Camera
• a convolutional neural network (CNN) built in PyTorch

The idea was to:

1. Capture live images during descent
2. Filter out blurry or tilted images using IMU data
3. Correct the perspective
4. Run them through a neural network
5. Build a heat map of suitable landing areas

It was an early exposure to computer vision and embedded ML-long before “edge AI” became mainstream.

3. Attempted Tertiary Mission: Guided Landing#

Initially, we even wanted to steer the payload toward the best landing site using a parafoil parachute controlled by servos. This required:

• modeling lift-to-drag ratio
• calculating stable glide paths
• designing multi-line parafoil control
• simulating wind influence

While the idea was scientifically solid, reality reminded us that engineering is hard. After multiple failed prototypes (and many tangled parachutes), we ultimately dropped the guided-landing functionality and switched to a reliable round parachute.

System Architecture#

Our CanSat was effectively a two-computer system:

Microcontroller (ATSAMD21G18)#

• Reads temperature & pressure
• Sends telemetry every second via LoRa
• Stores data on microSD
• Runs the entire primary mission independently

Raspberry Pi Zero#

• Captures images
• Processes IMU + GPS data
• Runs terrain classification (CNN)
• Generates landing heat maps
• Coordinates all secondary mission logic

The Pi communicated with its peripherals using:

• I²C for sensors
• UART for GPS
• Camera ribbon for PiCam
• SPI / SDIO for storage

Despite the tiny form factor, it was a surprisingly complete embedded system.

Mechanical Design#

Fitting a Raspberry Pi, camera, parachute mount, servos, and a full power system into a soda-can-sized shell was a miniature engineering challenge.

CanSat Case#

Designed in Autodesk Fusion 360, our case used:

• top and bottom plates
• 4 vertical guiding rods
• 6 external mounting holes
• 3D-printed internal supports

This modular design allowed fast inspections and repairs-something we quickly learned was essential.

Vertical Mount for Raspberry Pi#

We created a dedicated 3D-printed vertical frame to hold the Pi Zero and extension boards. Given the tiny footprint and strict volume limits, this was the only feasible orientation.

Recovery System Engineering#

This was my part of the project-and the steepest learning curve.

Prototype 1: Parafoil (Clark-Y Profile)#

Our first goal was to build a controllable parafoil with:

• non-zero lift-to-drag ratio
• left/right steering via servos
• predictable glide performance

We built it from nylon fabric with household strings as suspension lines. It deployed-but didn’t glide. It simply descended like a normal parachute.

We realized we had incorrectly adapted round-parachute formulas to a parafoil, which behaves completely differently aerodynamically.

Prototype 2: Parafoil (Based on Academic Research)#

We rebuilt everything using aerodynamic equations from:

• Om Prakash, Aerodynamics, Longitudinal Stability and Glide Performance of Parafoil/Payload System

This time we:

• adjusted aspect ratio
• optimized chord length & canopy span
• changed line thickness to 0.2 mm
• recalculated lift/drag forces

The updated parachute flew much better-but had problems:

• too large for the CanSat container
• inconsistent openings
• frequent line tangling
• delayed inflation

Ultimately, we determined the design wasn’t reliable enough for the competition constraints.

Final Design: Round Parachute#

Based on NASA and CanSat training materials, we switched to a round parachute-simple, reliable, predictable.

And most importantly: it fit inside the can.

Testing the System#

We tested each subsystem separately before integrating:

Primary Mission Tests#

We verified telemetry accuracy by reading:

• frame number
• pressure
• temperature

Comparing DS18B20 and BMP280 validated sensor accuracy.

Secondary Mission Tests#

Testing the parafoil prototypes took weeks:

• balcony drops
• school-building drops
• indoor glide tests
• multiple redesigns

While the neural network prototypes worked in simulation, real airborne testing was limited by our parachute delays.

Project Timeline & Iterations#

Between November 2018 and March 2019, we built:

• 4 main hardware prototypes
• 3 parachute generations
• 2 software stacks
• dozens of drop tests
• a full outreach program (blog, shirts, presentations)

It was more than a school project-it was a full engineering cycle.

Budget & Sponsors#

Total budget: ~500 PLN (~116 EUR) Most parts were sponsored by:

• TPU Sp. z o.o.
• BOTLAND (discounts and support)

Outreach & Public Engagement#

We documented the project with:

• a Facebook page
• technical blog posts
• school presentations
• visual branding & logo design
• even an article in local newspaper

It was a great introduction to communicating engineering work publicly.

Conclusion: What I Learned#

This project taught me more than any class at the time:

• how to work in a team
• how to design, test, and iterate hardware
• how to integrate software with real sensors
• how to model aerodynamic systems
• how to manage complexity under tight constraints
• how to fail, recover, and move forward

“I Believe I Can Land” was my first real multidisciplinary engineering experience. It sparked my interest in embedded systems, physics-based modeling, and designing things that interact with the real world-a theme that continues in my later projects.