Introduction
We’re the Club Robot of INSA Rennes, part of the student association AEIR (Association of the students of INSA Rennes). Founded in 1998, we’ve been competing in the Robotic’s French cup ever since. This event organised by Eurobot and Planète des Sciences is one of the largest gatherings of engineering students in Europe, drawing over 200 teams from universities and engineering schools all across France and even further.
Our team is made up of around 15 students from different fields such as computer sciences, electronics, mechanical engineering, telecommunications, and more. Some of us are first-year students learning the ropes; others are in their final year. That mix is what makes the club work indeed knowledge transfer is at the heart of what we do.
Our project
Each year, the Robotic’s French cup introduces a brand-new theme. In 2026, the scenario revolves around squirrels racing to hoard hazelnuts before winter. Autonomous robots are competing on a shared game board, scoring points in 100 second matches without any human interferences. The main robot needs to grab sets of planks. One side is yellow the other is blue, in order to score, the planks needs to be positioned facing the team’s color at specific locations marked on the table.
Our electronics
Electronics are at the core of our robot, and this year we pushed both the
architecture and the hardware further than ever before.
The brain: STM32H7
The main control board is built around an STM32H7. It is a powerful Cortex-M7
microcontroller that handles high-level decision making, strategy, and coordination
across all subsystems. It acts as the master in our distributed architecture,
orchestrating everything from locomotion to the gripper system.
A modular motherboard architecture
One of our big electronics goals this year was reusability. We designed a
motherboard system that accepts interchangeable modules, which are small PCBs
that plugs in and expose a standardised set of interfaces : I2C, UART, analog pins,
and more. Depending on how many pins a function requires, a module can be single,
double, or triple width.
- A double module managing the connection with the gripper system,
- A module dedicated to the HMI .
The gripper: A 4-finger design with per-finger PCBs
Our main gripper this year is split into 4 independent “fingers”. Each finger needs to
be smart They carry:
- A color sensor to identify game elements,
- A servo motor to rotate and orient the wooden board which are being grabbed,
- A current-controlled DC motor to grip with consistent and repeatable force.
To wire all of this cleanly, we designed a small dedicated PCB for each finger. All
four finger boards are daisy-chained in series, which dramatically reduce cable
count and routing complexity inside the gripper. Each finger PCB hosts an STM32C0
microcontroller, which handles local sensor reading and motor control, and
communicates with the main STM32H7 over the chain.
This distributed approach, allows the master STM32H7 delegating to per-finger
STM32C0s. This keeps the main controller free for high-level tasks while giving
each finger its own responsive, low-latency control loop.
Our favorite Aisler features
Three things stand out for us.
First, delivery speed. Being based in Rennes, France, the geographical proximity to
Aisler means that our boards arrive fast which matters a lot when you’re iterating
under competition deadlines.
Second, the Blitz service has been a genuine lifesaver. When you’ve just spotted a
bug and need to re-spin a board quickly, knowing you can get PCBs back in days
rather than weeks changes how you work. It gives you the confidence to iterate
rather than second-guess every design decision.
Third, the KiCad integration fits naturally into our workflow. No export headaches,
no format gymnastics, just a smooth path from schematic to order.
Something that went wrong,the I2C address disaster
The first version of the finger PCB was ambitious. Rather than putting a
microcontroller on each finger, we designed a more “lightweight” approach: a PWM
controller over I2C to drive the motor driver, as well as an I2C ADC to read motor
current consumption through the driver.
The problem? Our color sensors don’t have a configurable I2C address. With four
identical sensors on the same bus, we needed address translation, so we added a
dedicated I2C address translator to the design.
We ordered the PCBs through Aisler. The boards came back fast and looking great.
Then, while waiting for the component order to be fulfilled, we received an email:
"Due to restrictions imposed by the manufacturer, we are unable to ship the
following product." The address translator was simply unavailable for export.
Lesson learned: always verify component availability and export restrictions before
finalising your BOM, not after you’ve already ordered the PCBs.
But the story has a happy ending. The redesign forced us to rethink the architecture
entirely, and the STM32C0-based solution turned out to be strictly better: the I2C
ADC we had originally chosen had a refresh rate too low to implement reliable
current-based grip control. The C0 gave us the sampling speed we needed for
consistent, repeatable gripping. Sometimes a parts shortage is the best design
review you’ll ever get.