A Python-based robotic control system and graph-search navigation pipeline for autonomous rovers.
The challenge was to program a robot to explore a randomly generated maze until it received a specific destination coordinate from a central server. The robot then had to navigate to this server-assigned target efficiently while adhering to strict memory and timing constraints.
We engineered a modular Python codebase centered around a finite state machine. We implemented a robust communication layer that handled MQTT messages to parse the target coordinates. We used a mapping algorithm, specifically depth-first search, to explore the maze and a pathfinding algorithm to calculate the route to the server-assigned target once received.
The robot successfully passed the final examination under tight time pressure, correctly receiving the server payload, mapping the maze, and navigating to the destination without error or memory leaks.
The software runs on the ev3dev Linux kernel. It utilizes a main event loop that polls sensors, including color, distance, and gyro sensors, and feeds data into a central logic controller. An asynchronous MQTT client handles the negotiation with the server to receive the target coordinates.
Logic to request, parse, and validate target coordinates received from the central server via MQTT.
Implemented Proportional-Integral-Derivative logic for smooth line tracing and wall alignment.
Robust logic handling for switching between exploration and target navigation modes.
Algorithms to map the maze structure in memory and calculate the shortest path to the assigned goal.
Calibrated edge sensors and established telemetry sessions via MQTT.
Developed state structures and depth-first exploration loops.
Refined routing algorithms to calculate paths rapidly.
rover successfully completed the formal evaluation under real-time constraints.