What are Automated Guided Vehicles?
Automated Guided Vehicles are self-moving machines designed to transport materials or
products in warehouses and manufacturing environments. Equipped with sensors, cameras,
and control software, they navigate, detect obstacles, and follow planned routes with
minimal human intervention. Our prototype is a small, cost-effective AGV built with
a stereo camera — designed for small businesses and light industry, delivering high
performance at a fraction of the cost of traditional systems.
AGV navigation methods
The right navigation approach depends on the application requirements and operating environment. These are the five most common methods:
🔴Laser-guided navigation
Uses lasers to detect reflectors positioned along the path — determining location and direction with high accuracy in structured environments.
🧲Magnetic guidance
Follows magnetic tape or embedded floor wires — simple, low-cost, but limited to fixed predefined routes that cannot change without floor modification.
📷Vision-based navigation
Employs cameras and image recognition to detect floor markings, landmarks, or environmental cues — enabling more flexible routing and dynamic adaptation.
📡Inertial navigation
Relies on acceleration and rotation sensors to estimate position and movement — useful in environments where external references are unavailable.
🏷️RFID navigation
Reads RFID tags embedded in the floor to determine location — precise position identification at predefined points with minimal infrastructure.
Sensor technologies: LiDAR vs stereo cameras
Two of the most effective sensor technologies for AGV navigation — and they work best when used together.
📡 LiDAR sensors
Laser beams measuring distances to create accurate 2D or 3D maps
Highly effective for detecting walls, machinery, and shelving obstacles
Reliable in low light and dusty environments
Limited colour and texture information
📷 Stereo cameras
Two lenses capturing images from different perspectives for depth perception
Rich visual information — colour, texture, and detailed 3D environmental mapping
Enables visual landmark recognition and CNN-based object detection
Combined with LiDAR: improved accuracy, obstacle avoidance, and environmental awareness
ROS: the backbone of autonomous navigation
ROS (Robot Operating System) is the open-source framework that underpins most
autonomous robot navigation. Its Navigation Stack provides all components needed for
fully autonomous operation — from mapping to motion control.
1
Localisation
Determines the robot’s current position using sensors (LiDAR, cameras) and SLAM — Simultaneous Localization and Mapping — building a map while tracking position within it.
2
Mapping
Builds and maintains a map of the operating environment from sensor data — either using a preloaded map or constructing one on the fly as the AGV explores.
3
Path planning
Calculates the optimal route from current position to target destination while avoiding all detected obstacles — replanning dynamically when the environment changes.
4
Motion control
Executes the planned path — controlling speed, steering, and movement precision — ensuring smooth, accurate navigation to the target location.
AI transforming AGV navigation
AI moves AGVs from rigid, pre-programmed systems to adaptive, self-learning platforms —
capable of navigating environments that change without requiring human reprogramming.
Deep learning — processing complex sensor data to detect obstacles, landmarks, and paths with CNN-based visual recognition
Reinforcement learning — training AGVs to choose optimal routes through trial-and-error in simulated and real environments
Continuous learning — improving from previous runs, adapting to layout changes without manual reprogramming
Predictive maintenance — forecasting maintenance needs and detecting early signs of malfunction to reduce downtime
Benefits of AI-powered AGV navigation
✓
Dynamic environment navigation — AGVs adapting to layout changes and unexpected obstacles in real time
✓
No fixed-route infrastructure — no magnetic tape, floor wires, or reflectors required
✓
On-the-fly map building — SLAM enabling operation in new or changed environments from first deployment
✓
Continuous improvement — AI getting better with every route without manual model updates
