GNSS (global navigation satellite system) is the umbrella term for satellite constellations that broadcast precisely timed radio signals from which a receiver computes its absolute position, velocity and time anywhere on Earth with open sky. In robotics navigation, GNSS gives outdoor robots such as agricultural vehicles, delivery robots, drones and survey platforms a global reference frame that does not drift over time.
Four global systems are operational: the American GPS, the European Galileo, the Russian GLONASS and the Chinese BeiDou, supplemented by regional systems such as Japan’s QZSS and India’s NavIC. Modern multi-constellation, multi-frequency receivers track several of them at once, which improves availability and accuracy in difficult environments.
How GNSS positioning works
Each satellite transmits a signal that encodes its orbit data (ephemeris) and the time of transmission from an onboard atomic clock. The receiver measures when the signal arrives and multiplies the travel time by the speed of light to obtain a pseudorange, a distance that still contains the receiver’s own clock error. With at least four satellites, the receiver solves for three position coordinates and its clock offset.
Accuracy is limited by several error sources: ionospheric and tropospheric delays, satellite clock and orbit errors, multipath reflections from buildings or terrain, and receiver noise. The geometry of visible satellites amplifies these errors, which is expressed by the dilution of precision (DOP); values below about 2 indicate good geometry, while values above 5 indicate weak geometry.
The table below lists the main error sources and how they are typically reduced.
| Error source | Effect | Typical mitigation |
|---|---|---|
| Ionospheric delay | Several meters of range error | Dual-frequency measurements, corrections |
| Tropospheric delay | Range error rising at low elevation | Models, elevation mask |
| Satellite orbit and clock | Meter-level errors | Precise correction products |
| Multipath | Jumps and biases near reflecting surfaces | Antenna design, placement, signal processing |
| Poor geometry | Amplified position error | More constellations, higher DOP awareness |
| Obstruction | Too few satellites or no fix | Sensor fusion with odometry and inertial sensors |
Accuracy levels and correction methods
Positioning accuracy depends far more on the correction method and receiver type than on the constellation itself. A standalone single-frequency receiver typically achieves a few meters, while carrier-phase methods reach centimeter-level accuracy.
| Method | Typical horizontal accuracy | Requirement | Robotic use |
|---|---|---|---|
| Standalone single-frequency | About 2 to 5 m | Basic receiver | Coarse geofencing, tracking |
| Standalone dual-frequency | About 1 to 2 m | Multi-band receiver | Delivery robots with other sensors |
| SBAS | About 1 m | Regional augmentation signal | Drones, general navigation |
| DGNSS | Decimeters to 1 m | Nearby reference correction | Marine and field vehicles |
| PPP | Centimeters to decimeters after convergence | Precise satellite corrections, minutes to converge | Remote areas without base stations |
| RTK | About 1 to 2 cm | Base station or network within tens of kilometers | Agriculture, surveying, row-following robots |
RTK GPS is the common name for real-time kinematic positioning, which uses carrier-phase measurements and corrections from a nearby base station. RTK requires resolving integer ambiguities; the solution status is reported as fixed (centimeter level) or float (decimeter level), and a robot should treat these states differently.
GNSS in robot localization
Robots rarely navigate on GNSS alone. GNSS updates arrive at 1 to 20 Hz, may jump or disappear under trees and near buildings, and provide no heading when the robot is stationary. Robots therefore fuse GNSS with wheel odometry, inertial measurement units and sometimes lidar in an extended Kalman filter or a factor graph.
In ROS-based systems, a common setup transforms GNSS latitude and longitude into a local metric frame and feeds it to a state estimator together with odometry. The estimator weighs each GNSS fix by its reported covariance, so a poor fix has little influence. Dual-antenna receivers add an absolute heading, which is valuable for slow-moving vehicles.
| Source | Reference frame | Strength | Weakness |
|---|---|---|---|
| GNSS | Global, absolute | No long-term drift | Needs open sky, vulnerable to interference |
| Wheel odometry | Relative | Smooth, high rate | Drifts with slip and wheel wear |
| Inertial measurement unit | Relative | Very high rate, independent of environment | Drift grows quickly over time |
| Lidar or visual localization | Map-relative | Works indoors, high precision | Needs a map and recognizable features |
| Dead reckoning | Relative | Bridges short outages | Error accumulates with distance |
Choosing a GNSS setup for a robot
The required accuracy should drive the setup. A lawn robot following a boundary within a few centimeters needs RTK with a fixed solution, while a sidewalk delivery robot can use meter-level GNSS combined with visual localization. Antenna quality and placement often matter more than the receiver brand.
| Application | Needed accuracy | Suitable setup |
|---|---|---|
| Agricultural row following | About 2 cm | RTK with network or local base, IMU fusion |
| Lawn and turf robots | A few centimeters | RTK base at the site, boundary map |
| Sidewalk delivery | About 1 m | Multi-band GNSS fused with cameras and odometry |
| Inspection drones | Centimeters to meters | RTK or PPK for mapping, SBAS for flight |
| Asset tracking robots | Several meters | Standalone GNSS |
Practical rules include mounting the antenna high and clear of metal structures, using a ground plane, keeping cable runs short and setting an elevation mask of about 10 to 15 degrees to discard noisy low satellites.
Limitations and common pitfalls
GNSS signals are extremely weak at the Earth’s surface, so they are easily blocked, reflected or disrupted. Jamming and spoofing are real threats for autonomous systems, and safety-critical robots should detect implausible jumps and fall back to other sensors.
- Trusting float RTK as if it were a fixed solution.
- Ignoring datum and coordinate frame differences, which can shift positions by meters.
- Placing the antenna where the robot body causes multipath.
- Not handling outages in urban canyons and under dense canopy.
- Forgetting the lever arm between antenna and robot reference point.
Understanding it simply
GNSS is the family of satellite systems, including GPS, that lets a device work out where it is on Earth. Satellites keep sending out the exact time, and a receiver calculates its distance to several of them from how long the signals took to arrive.
For a robot outdoors, GNSS works like a map pin that never wanders off over time. It can be blocked by buildings or trees, so robots combine it with their own wheel and motion sensors to stay on course.
Applied in practice
A vegetable farm uses an autonomous weeding robot that drives between crop rows spaced 30 centimeters apart. A standalone receiver would place it a meter or more off course, which would destroy plants.
The robot uses a multi-band RTK receiver with corrections from a base station at the farm, fused with wheel odometry and an inertial unit. When the RTK status drops from fixed to float near a tree line, the robot slows down and relies on camera row detection until the fixed solution returns. It holds its path within a few centimeters across fields of several hectares.