2023 · Hardware / Robotics
Maze-Solving Autonomous Robot Car
Graph-theory driven autonomous maze navigation on a Raspberry Pi Pico.
TVSEF 2023 Award Winner · Graph Theory
The problem
Navigate a maze you've never seen with no GPS and no LiDAR. The car has to walk the walls, remember every cell it maps, and recover from dead ends — all with cheap binary sensors and a microcontroller.
The constraint
The Pi Pico reads forward/left/right/down from distance sensors that bounce unpredictably off angles. Its motors have no encoders, so the car can't trust odometry: it must *calibrate* how far it rolls and infer its position cell-by-cell from sensor events, not from wheel counts.
What I built
The project proves the graph algorithm *before* the hardware. In Pygame, a simulation loads a maze from a text file, runs a breadth-first search from start to goal, and animates the path (stack overlays, visited nodes) over a grid. On the real Pi Pico, MicroPython runs the same idea in three acts: a calibration step measures how long 447 mm of travel takes to derive a block-timing constant, a wall-follow loop drives the maze with the four sensors and a dead-end recovery (buzz and reverse), and a mapping routine builds the collected maze dynamically in memory, growing its grid as the car enters new rows/columns.
The outcome
Demonstrated autonomous exploration, dead-end recovery, and solved-maze traversal on physical hardware, backed by a simulation that proved the algorithm — winning a TVSEF 2023 regional science fair award.