Building AGVs: QR Navigation & Lifter Systems

Learning Goal: Build an industrial-style AGV (Automated Guided Vehicle) that navigates a grid layout using a downward-facing camera for QR code landmark localization and a lifter mechanism.

Estimated Total Study Time: 32 hours

Module 1: Introduction to AGVs & Embedded Electronics

This module provides a comprehensive foundation for understanding how industrial automated guided vehicles operate in modern factories. You will master the physics of differential drive kinematics and study the hardware configuration of high-power DC motor drivers and microcontrollers needed to move heavy physical payloads.

Knowledge Checkpoint

Calculate the forward velocity ( $v$ ) and angular velocity ( $\omega$ ) of a differential drive robot given the wheel radius ( $r$ ) and individual wheel angular speeds ( $\omega_l$ , $\omega_r$ ).
Connect the BTS7960 motor driver to an Arduino or Raspberry Pi and configure the enable pins (R_EN, L_EN) and PWM pins (R_PWM, L_PWM) to achieve bidirectional speed control.
Trace the signal path from a path-planning microcontroller to the motor driver, identifying potential noise isolation issues and power ground loop hazards.

Module 2: Chassis Design & Lifter Mechanics

AGVs must not only navigate, but also interact with physical payloads. This module guides you through the CAD design process of a mobile chassis and dives into the mechanical engineering principles of lifting mechanisms—specifically scissor lifts and linear actuator lead screws.

Physical Structural Integration Guidance

Curriculum Note: When integrating your lifting mechanism with a mobile chassis:

Ensure the center of gravity of the payload remains directly centered between the drive wheels to prevent tipping.

Mount the linear actuator base directly to the structural frame plates of your chassis rather than relying on thin 3D-printed paneling.

Route all lift-mechanism power cables through dynamic cable tracks to prevent binding during compression.

Knowledge Checkpoint

Draw a free-body diagram of a one-stage scissor lift mechanism and identify the point of maximum mechanical stress during structural extension.
Determine when to choose a multi-start lead screw over a single-start threaded rod based on mechanical efficiency and self-locking requirements.
Design structural brackets to mount a linear actuator to your AGV's lower chassis plates while ensuring proper clearance for the drive motors and battery packs.

Module 3: Computer Vision & QR Code Localization

This module teaches you to turn a downward-facing camera into a sub-millimeter precision localization sensor. You will master OpenCV-based QR code extraction and the geometric homography equations necessary to convert pixel coordinates into precise real-world coordinate offsets on your warehouse grid floor.

Camera-to-Ground Homography Implementation

To transform a detected QR code corner in pixels $[u_i, v_i, 1]^T$ to real-world floor offsets $[x_i, y_i, 1]^T$ relative to the camera center:

\begin{bmatrix} x \\ y \\ 1 \end{bmatrix} \sim H \begin{bmatrix} u \\ v \\ 1 \end{bmatrix}

Where $H$ is a $3 \times 3$ Homography matrix computed during system calibration using 4 known points on your floor grid. Once $H$ is stored on the AGV's companion computer (e.g., Raspberry Pi), any QR detection yields an instant centimeter-level error offset vector ( $\Delta x, \Delta y$ ) and heading error ( $\Delta \theta$ ).

Knowledge Checkpoint

Calibrate your camera using a checkerboard pattern to find its focal length ( $f_x, f_y$ ) and principal point ( $c_x, c_y$ ) values.
Write a Python script using OpenCV (cv2.findHomography) to calculate the homography matrix $H$ mapping four pixel locations to four physical points.
Explain how lens distortion parameters ( $k_1, k_2$ ) affect the coordinate accuracy of landmarks detected near the edges of your camera’s field of view.

Module 4: Grid Navigation & Motion Control Algorithms

With localization working, this module shows you how to close the feedback loop. You will program PID controllers to smoothly align the vehicle over landmarks and implement a robust Finite State Machine (FSM) to coordinate movement, alignment, and lifter activation.

Python Finite State Machine (FSM) Template

For an AGV to coordinate traveling, landmark alignment, and lifter activation, organize your script around this standard state execution model:

class AGVStateMachine: def init(self): self.state = "IDLE" # Initial state self.target_qr = None

def spin_once(self, camera_data, odometer_data):
    if self.state == "IDLE":
        # Wait for dispatch command
        self.target_qr = self.get_next_job()
        if self.target_qr:
            self.state = "NAVIGATING_GRID"

    elif self.state == "NAVIGATING_GRID":
        # Run path-following control loop
        if camera_data.qr_detected == self.target_qr:
            self.state = "PRECISE_ALIGNMENT"
        else:
            self.send_drive_commands(self.calculate_grid_path())

    elif self.state == "PRECISE_ALIGNMENT":
        # Engage homography-based PID centering loop
        error_x, error_y, error_theta = camera_data.get_homography_offsets()
        if abs(error_x) < 0.005 and abs(error_y) < 0.005:  # 5mm tolerance
            self.stop_motors()
            self.state = "ACTIVATE_LIFTER"
        else:
            self.send_drive_commands(self.calculate_pid_velocity(error_x, error_y, error_theta))

    elif self.state == "ACTIVATE_LIFTER":
        # Trigger physical lift payload
        if self.engage_lifter_mechanism(direction="UP"):
            self.state = "IDLE"  # Action completed!

Knowledge Checkpoint

Implement the provided Python FSM framework on your companion computer to handle state transitions from standard grid navigation to precise alignment over a landmark.
Tune a positional PID loop's proportional gain ( $K_p$ ) to achieve rapid convergence over a QR code target without causing the robot to oscillate.
Implement windup protection limits (clamping) for the Integral term ( $K_i$ ) of your position and rotation controllers.

Module 5: System Integration, Safety, & Testing

This module pulls all the mechanical, electrical, and computational components together. You will integrate LiDAR and other safety sensors, set up physical mock warehouse courses, and perform diagnostic integration tests to verify the AGV runs reliably under standard operating conditions.

Safety System Field-of-View Integration

To protect operators and payloads, map your safety sensors (LiDAR/Ultrasonic) directly to override interrupts inside your motor control loop:

              +----------------------------------+
              |  LiDAR / Distance Sensors Scan   |
              +-----------------+----------------+
                                |
                Is obstacle inside Warning Zone?
                /                              \
             YES                                NO
             /                                    \

Knowledge Checkpoint

Connect a 2D LiDAR or array of distance sensors to your companion computer and map their ranging profiles to spatial zones relative to your chassis coordinates.
Program a hardware interrupt that immediately cuts off enable signals to the BTS7960 motor driver if an obstacle is detected within 30 centimeters of the robot.
Conduct a diagnostic integration test: log your camera-to-ground alignment error over 20 repeated approaches to confirm positional repeatability stays within a 5-millimeter window.

Course Map

Key People Index

Prof. Peter Corke (Professor of Robotic Vision, QUT)
- Context: Featured in Module 3. Author of the foundational textbook Robotics, Vision and Control. Corke provides the definitive mathematical proofs for planar transformations and image projective geometry.
Prof. Cyrill Stachniss (University of Bonn)
- Context: Featured in Module 3. A world-renowned researcher in SLAM and photogrammetry. Stachniss provides highly regarded, open lectures on camera calibration algorithms and geometric sensor models.
James Bruton (Robotics Designer & Educator)
- Context: Featured in Modules 4. Bruton is a former toy designer and electrical engineer known for physical prototyping, hardware-in-the-loop testing, and implementing accessible state machine patterns.

Final Self-Assessment

Perform this final diagnostic checklist to verify your finished AGV platform meets all core technical requirements:

Power Safety: The power delivery network utilizes an inline fuse between the battery and the BTS7960 motor driver to protect components from overcurrent situations.
Kinematic Control: The AGV moves forward 1.00 meter on a level floor with less than 2 centimeters of lateral drift without active sensor feedback.
Robust Calibration: The downward-facing camera has successfully generated its camera matrix and distortion coefficients using Zhang's checkerboard method.
Coordinate Homography: The vision pipeline detects a floor-mounted QR code and calculates physical coordinate offsets ( $x, y, \theta$ ) relative to the camera's optical center.
State Machine Management: The FSM controls state transitions from path-following navigation to precise QR alignment, then to lifter engagement without getting stuck in a loop.
Physical Lift Performance: The vertical lift mechanism can successfully raise and support a 2-kilogram payload without binding, slipping, or drawing excessive current.
Safety Override: The obstacle avoidance sensors (LiDAR/Ultrasonic) override active drive commands and bring the vehicle to a stop if a person or obstacle blocks its path.
Target Repeatability: The integrated system can navigate from an idle state, locate a target QR code on the floor, align itself, lift the target payload, and move to a drop-off point autonomously.

Explore Further

Related Robotics Roadmaps

View All→

robotics

Building AGVs: QR Navigation & Lifter Systems

Module 1: Introduction to AGVs & Embedded Electronics

Recommended Videos

Knowledge Checkpoint

Module 2: Chassis Design & Lifter Mechanics

Recommended Videos

Physical Structural Integration Guidance

Knowledge Checkpoint

Module 3: Computer Vision & QR Code Localization

Recommended Videos

Camera-to-Ground Homography Implementation

Knowledge Checkpoint

Module 4: Grid Navigation & Motion Control Algorithms

Recommended Videos

Python Finite State Machine (FSM) Template

Knowledge Checkpoint

Module 5: System Integration, Safety, & Testing

Recommended Videos

Safety System Field-of-View Integration

Knowledge Checkpoint

Course Map

Key People Index

Final Self-Assessment

Related Robotics Roadmaps

CDPR: Kinematics, Arduino & Winch Control

Rocker-Bogie Rover: Build, Stereo Vision & ROS

EMG Exoskeletons: Design, Control & Actuation

VR Digital Twins: Unity, ROS & Teleoperation