Autonomous Fire-Extinguishing Robot

An autonomous, holonomic mobile robot engineered for real-time environment navigation, dynamic obstacle avoidance, optical flame localization, and target extinguishing. Developed as part of the MECHENG 706 (Mechatronics Systems) curriculum at the University of Auckland.

Developers:

• Abhijot Malhi — @tabyrocket
• Wayne Liang — @catxna

Achievement: This robot was the only entry in the cohort to achieve a 100% score across all benchmark evaluation runs, demonstrating complete reliability in flame detection speed, navigation accuracy, obstacle clearance, and target extinguishing.

---

📷 Robot Overview

🎬 Autonomous Benchmark Run Demo

Click to view the full autonomous run video: figures/RobotTestRun.mp4

---

🧠 Core Engineering & Systems Architecture

graph TD
    S1[State 1: Initialisation & Calibration] --> S2[State 2: 360° & Swept Flame Search]
    S2 --> S3[State 3: MIMO Fuzzy Navigation & Flame Tracking]
    S3 -->|Obstacle Trap / Wheel Stall| S6[State 6: IMU Stall Recovery & Escape]
    S6 --> S3
    S3 -->|Target Acquired & In Range| S4[State 4: Closed-Loop Aim & Extinguish]
    S4 -->|Flame Verified Extinguished| S5[State 5: Safe Disengage & Reverse]
    S5 -->|Fires Remaining > 0| S1
    S4 -->|All Targets Extinguished| END[Mission Complete / Standby]

Loading

1. 4-Wheel Holonomic Drive & Kinematics Engine

The chassis utilizes a 4-wheel omnidirectional configuration to achieve unconstrained 3-DOF planar motion (independent translation in $x, y$ and rotation $\omega$).

• Inverse Kinematics Formulation: Parameters: Wheel radius $R_w = 30,\text{mm}$, half-wheelbase length $L = 105,\text{mm}$, half-track width $l = 90,\text{mm}$.
• Decoupled Kinematics (moveDirectionCapX): Preserves prioritized lateral avoidance ($y$) and angular correction ($\omega$) by dynamically recalculating forward velocity headrooms within $[\pm 500]$ motor PWM bounds.
• Dynamic Drift Compensation & Deadband Handling: Closed-loop non-linear friction break-away thresholds ($\ge 85,\text{PWM}$) and heading cross-coupling compensators ensure straight-line fidelity.

---

2. Multi-Input Multi-Output (MIMO) Fuzzy Logic Controller

Reactive obstacle avoidance and path planning are governed by a C++ fuzzy inference system (FuzzyLogic) calculating exact analytical centroids.

• Fuzzification: 5 input linguistic variables across 17 membership sets:
  • Ultrasonic Front Distance ($Z, S, \text{Triangular}$)
  • Left / Right Short-Range IR ($Z, S, \text{Triangular}$)
  • Front-Left / Front-Right Long-Range IR ($Z, S, \text{Triangular}$)
  • Gyro Heading Error & Phototransistor Differential Corrections
• Analytical Defuzzification: Real-time geometric centroid computation $\left(\frac{\sum u_i \cdot A_i}{\sum A_i}\right)$ evaluating exact trapezoid/triangle integration without discrete numerical lookup tables.
• Rule Aggregation: Smoothly balances aggressive lateral wall bypassing against continuous optical flame attraction.

---

3. Sensor Fusion & Digital Signal Processing

• BNO08x 9-DOF Intelligent IMU: Tracks continuous 3D game rotation vectors (SH2_GAME_ROTATION_VECTOR) with quaternion-to-Euler yaw extraction and phase unwrapping for drift-free heading reference.
• 4-Channel Infrared Rangefinder Array: Combines analog GP2Y0A21YK0F and GP2Y0A41SK0F sensors calibrated with power-law voltage-to-distance transforms: $$\text{Distance} = C \cdot V^{\gamma} - d_{\text{shift}}$$
• Phototransistor Target Array: 4-channel analog optical detector array configured for flame triangulation and differential tracking.
• First-Order Recursive Low-Pass Filtering: $$y_k = y_{k-1} + K \cdot (x_k - y_{k-1}), \quad K = \frac{P}{P + Q}$$ Mitigates noise and ambient light flicker across all high-speed telemetry channels.

---

4. Robust 6-Stage Finite State Machine (FSM)

The autonomous executive loop (FSM) provides deterministic state handling with built-in fault tolerance:

State	Name	Functional Description
1	Initialisation	Sensor zeroing, digital filter seeding, and IMU baseline gyro calibration.
2	Search	Bidirectional $150^\circ$ servo sweep coupled with $120^\circ$ yaw rotations to detect optical flame signatures. Rejects false-positives via peak irradiance filtering.
3	Move / Track	Fuses MIMO fuzzy avoidance with differential flame heading vectors. Includes a persistent lateral wall-bypass latch for escaping concavities.
4	Extinguish	Motion-locks the drive base, performs closed-loop pan servo alignment, conducts safety proximity confirmation, and activates the high-flow fan until optical feedback confirms extinction ($< 10$ ADC threshold).
5	Reverse	Controlled recoil and sensor re-sampling maneuver prior to searching for secondary targets.
6	Stall Recovery	Detects stationary IMU yaw patterns during commanded turns/translations; executes multi-axis un-wedging sequences (lateral strafe + reverse).

---

🛠️ Hardware Specification & Pin Interface

Subsystem	Component	Interface / Pin Mapping
MCU	Arduino Mega 2560 (ATmega2560)	Main On-Board Controller
Drive Motors	4x Continuous DC Geared Motors	Pins 46 (FL), 51 (FR), 47 (RL), 50 (RR)
Extinguisher	High-RPM Centrifugal Fan + Servo Mount	Fan Pin 45 (PWM/Digital), Servo Pin 9
IMU	BNO080 / BNO085 9-DOF Sensor	I2C (SDA / SCL, 100 Hz SH2 Reports)
Ranging (Sonar)	HC-SR04 Ultrasonic Sensor	Trigger: 48, Echo: 49
Ranging (IR)	4x Sharp Distance Sensors (Long/Short)	Pins A8 (FL), A9 (L), A10 (R), A11 (FR)
Optical Sensors	4x Phototransistor Array	Pins A2 (Far-L), A3 (Mid-L), A6 (Mid-R), A5 (Far-R)
Diagnostics	RGB Status Indicator + Bluetooth UART	LEDs Pin 13 / Pin 12 / Pin 11, Serial Bluetooth

---

📁 Repository Structure

├── fire-extinguishing-robot.ino   # Main embedded setup and loop entry point
├── figures/                       # Demonstration videos, photos, and performance plots
│   ├── RobotMainView.jpg          # Isometric front photograph
│   ├── RobotBackView.jpg          # Rear subsystem photograph
│   ├── RobotTestRun.mp4           # 100% benchmark evaluation run recording
│   └── fuzzy_membership_*.png    # Generated fuzzy logic membership curves
└── src/
    ├── Actuators/                 # Actuator drivers (Motors, Fan, FanServo)
    ├── Sensors/                   # Sensor abstractions (BNO08x Gyro, IR, Sonar, Phototransistors)
    ├── Robot/
    │   ├── Control/               # Kinematics, trajectory generation, and aiming routines
    │   │   └── Fuzzy/             # Analytical MIMO Fuzzy Logic Inference Engine
    │   ├── FSM/                   # 6-State Autonomous Finite State Machine & Stall Watchdog
    │   └── Robot.cpp              # Hardware aggregation layer
    └── Misc/                      # Fast math helpers, digital filters, and telemetry tools

---

🚀 Building & Flashing

1. Prerequisites:
   • Arduino IDE or PlatformIO / Arduino-CLI
   • Target Board: Arduino Mega 2560
   • Libraries: Adafruit BNO08x, Wire, Servo
2. Compile and Upload:

   # Using Arduino CLI:
   arduino-cli compile --fqbn arduino:avr:mega fire-extinguishing-robot.ino
   arduino-cli upload -p <PORT> --fqbn arduino:avr:mega fire-extinguishing-robot.ino

3. Telemetry & Calibration: Open the Bluetooth/Serial terminal at 115200 baud to view real-time state machine transitions, filtered IR/Sonar readings, and PID/Fuzzy output telemetry.

---

🏆 Competition & Benchmark Performance

• Course: MECHENG 706 (Mechatronics Systems), Department of Mechanical & Mechatronics Engineering, The University of Auckland.
• Outcome: 100% Benchmark Completion Rate across all obstacle configurations and flame placement variations.
• Key Differentiator: The combination of true holonomic kinematics, custom continuous centroid fuzzy logic, and real-time optical closed-loop extinguishing eliminated blind spots and prevented deadlocks common in conventional differential-drive solutions.