Imperium is a fully autonomous chessboard that detects and moves pieces using a precision electromagnetic gantry, per-square Hall-effect sensing, and CNC-style motion control. The system integrates custom mechanical design, multi-PCB electronics, and firmware-level motion planning to deliver seamless, repeatable gameplay without manual calibration.
flowchart LR
Pieces[Chess Pieces — Magnets]
Sensors[Hall Sensor Matrix — 64 Squares]
Mux[16:1 Multiplexers]
MCU[ESP32 Master Controller]
Drivers[TMC2209 Motor Drivers]
Motors[Stepper Motors — T-Bot Gantry]
Magnet[Electromagnet]
Pieces --> Sensors
Sensors --> Mux
Mux --> MCU
MCU --> Drivers
Drivers --> Motors
MCU --> Magnet
Figure 2: Fully assembled Imperium chessboard
CAD Figure: T-Bot gantry layout and belt routing
Gantry Design
- Type: T-Bot (CoreXY derivative)
- Motors: 2× NEMA 17
- Linear motion: 2× MGN12 rails (450 mm)
- Drive: GT2 belts & pulleys
- End effector: Electromagnet (custom mount)
Design Rationale
- Compact footprint compared to CoreXY
- Sufficient precision for chess (no micron-level requirement)
- Reduced part count and cost
flowchart LR
Square1[Square 1]
Square2[Square 2]
SquareN[Square 64]
Mux1[16:1 MUX]
MCU[ESP32]
Square1 --> Mux1
Square2 --> Mux1
SquareN --> Mux1
Mux1 --> MCU
Figure 5: 64-square Hall-effect sensor grid
- Sensors: KTH-1601 (46 Gs threshold)
- 1 sensor per square (64 total)
- Mounted beneath ~3 mm acrylic
Why Hall Sensors
- Instantaneous response
- No ghosting (unlike reed switches)
- Reliable removal detection
flowchart TB
PCB1[Hall PCB Sheet 1\n16 Sensors]
PCB2[Hall PCB Sheet 2\n16 Sensors]
PCB3[Hall PCB Sheet 3\n16 Sensors]
PCB4[Hall PCB Sheet 4\n16 Sensors]
Master[Master PCB]
PCB1 --> Master
PCB2 --> Master
PCB3 --> Master
PCB4 --> Master
Figure 6: Hall sensor PCB sheet (100 mm × 400 mm)
Each PCB sheet contains:
- 16 Hall sensors
- 1× 16:1 multiplexer
- Local decoupling capacitors
- Ribbon cable interface
Total sheets: 4
Figure 7: Master PCB functional block diagram
Core Components
- MCU: ESP32
- Motor drivers: 2× TMC2209 (UART)
- Power:
- 12 V rail (motors + electromagnet)
- Buck converter → 3.3 V logic
- Electromagnet switching via transistor
All schematics and footprints were custom-designed in KiCad.
Chess pieces are moved exclusively along square edges rather than diagonals. This design choice:
Prevents collisions with stationary pieces
Simplifies motion planning
Ensures consistent and predictable movement paths
The gantry executes all moves as a sequence of orthogonal segments.
UART enables:
- Sensorless homing
- Stall detection
- Current tuning
- Advanced diagnostics
-
Board State Detection
Hall sensor matrix scans current piece positions. -
Move Translation
Chess moves converted into constrained paths. -
G-Code Generation
Custom G-code enforces non-diagonal movement. -
Execution
G-code executed by FluidNC on ESP32.
📎 Full BOM:
https://docs.google.com/spreadsheets/d/1yp7t6AiXMwJCAfVlhCX7udocGjks9lOOFWkv4rPJByo/view
- Reliable autonomous gameplay
- Accurate per-square detection
- Consistent homing and positioning
- No recalibration required between games
- Chess engine integration
- Companion UI / mobile app
- Faster traversal paths
- Noise reduction
- Fully enclosed consumer-ready housing
- CoreXY reference — https://corexy.com
- Gantry design pitfalls — https://drmrehorst.blogspot.com