Why There Is No One-Size-Fits-All Solution

First, the distinction. An AGV typically follows a predetermined route, often with the help of environmental cues such as magnetic strips or reflectors. An AMR is more autonomous: it can determine its own route, make course corrections, and avoid obstacles.

There are now proven solutions for a clean, predictable warehouse environment. But as soon as a robot has to operate in the real, messy world outside, those solutions stop working. Take a robot that has to navigate both indoors and outdoors: outdoors, a GNSS signal is available, but indoors it isn’t; lighting conditions vary greatly, as does the surface. Or consider a dirty, dusty environment like a livestock farm.

These kinds of circumstances give rise to two persistent problems:

• Poor visibility and pollution.
  Many modern navigation techniques rely on optical sensors: cameras and LiDAR (laser scanners). These work well in clean air with a clear line of sight, but can quickly run into problems due to dust and dirt. A cloud of dust kicked up by the wind can easily be mistaken for an obstacle, and a thin layer of dust or mud on the lens can blind the sensor. In such environments, cameras and LiDAR are therefore difficult to use or even unusable.
• Varying road surfaces and loss of traction.
  A robot that travels over various surfaces 
  is constantly dealing with varying levels of traction. Where the wheels slip or spin, traction is lost. This not only makes driving more difficult, but also undermines a commonly used method for estimating position. More on that later.

The conclusion: there is no one-size-fits-all solution. Every environment requires a carefully considered combination of techniques. It is precisely this customized approach—based on a view of the system as a whole—that we at RIWO excel at.

Positioning: The Foundation of All Navigation

Before a robot can travel even a single meter in a meaningful way, it must be able to answer this question: Where am I? Positioning is the foundation of navigation. A robot that doesn’t know where it is cannot determine where it needs to go or which route to follow. And it is precisely under the circumstances outlined above that this poses the greatest challenge.

The problem is that no single sensor is reliable everywhere on its own:

• GNSS, best known for GPS, uses satellites to estimate a position. With the RTK variant, it is possible to determine a position with an accuracy of just a few centimeters. GNSS works well outdoors, but loses signal completely indoors and becomes unreliable near buildings and structures.
• Optical sensors, including cameras and LiDAR, are often used in indoor environments. They provide accurate position determination when there is a clear line of sight, but it is difficult to ensure their reliability in dusty and contaminated environments.
• Wheel odometry—estimating position by counting how many revolutions the wheels make—is always available and inexpensive to implement. Its accuracy is reduced because tires behave differently on various surfaces. Loss of traction is disastrous and quickly leads to significant deviations. Furthermore, this technique is sensitive to tire wear.
• An IMU (Inertial Measurement Unit) is a type of sensor consisting of gyroscopes, accelerometers, and compasses. This allows the sensor to determine the robot’s orientation and how fast it is rotating. IMU measurements are always available and are not affected by traction loss, making it a popular choice for use in combination with wheel odometry. The built-in compasses are often unusable in indoor environments because they are easily disrupted by metal objects and electrical devices. An IMU can be used without compasses, but in that case, its readings will drift over time. Various techniques can be used to limit this effect.
• Landmarks in the environment, such as RFID (Radio Frequency Identification) tags embedded in the floor, can serve as an alternative to GNSS or optical sensors. Tags are installed at fixed, predetermined locations. When the robot drives over or past these tags, it reads them wirelessly and knows exactly where it is at that moment. Because this system relies on a radio signal and does not require a clear line of sight, it is particularly well-suited for dirty and dusty conditions. The limitation of this technique is that a measurement can only be taken when a tag is within range, and not at any points in between. This makes it difficult to establish a continuous, robust position based on this method.

The common thread: every sensor has its strengths, but there are also circumstances in which it performs less well or fails to provide any useful data at all. And those are precisely the circumstances that constantly arise in challenging environments.

Smart Combinations with Sensor Fusion

The solution does not lie in a single perfect sensor, but in the smart combination of different sensors. We achieve this through sensor fusion: combining measurements from different sensors into a single position estimate that is both more accurate and more robust than any single sensor on its own.

The idea behind sensor fusion is that each sensor not only provides a measurement, but also an indication of its reliability in a given situation. Sensor fusion weighs these measurements against one another and combines them to produce the best possible estimate of the robot’s location. Mathematically, this is achieved using techniques from statistics and control theory, such as Kalman filters.

The real advantage lies in its robustness. The sensors complement each other: if one sensor is less reliable in a given situation, the others make up for it. This way, the system doesn’t fail as soon as a single sensor malfunctions, but continues to provide a usable, reliable position. The result is a positioning system that adapts to changing conditions rather than being overwhelmed by them.

Comparing it to ourselves makes it easier to understand. People don’t navigate using their eyes alone either: we combine what we see with our sense of balance and the sensation of the steps we take. If you close your eyes for a moment, you can still walk straight ahead for a short distance without falling over. This is because your senses complement each other. A robot with sensor fusion essentially does the same thing.

From Position to Route

Once the robot can reliably determine its location, navigation comes into play: planning a route and following it accurately. At RIWO, we have extensive experience with both route planning and route following for ground robots.

Moreover, it’s rarely as simple as driving from point A to point B. Many applications don’t have fixed starting and ending points. For example, a feeding robot must first pick up feed, then drive to various locations to dispense measured portions of feed, and return to the charging station on time in between. That requires smart planning. Which tasks in what order? Which route is most efficient at that moment? And as soon as multiple robots are operating simultaneously, an additional challenge arises. How do you ensure that they don’t get in each other’s way and that their routes remain properly coordinated?

Faster Development Through Simulation

We first develop and test a large part of the positioning and navigation functionality using simulators. To do this, we use digital twins: virtual copies of the robot and its environment. In modern robotics, simulation has become an indispensable part of the development process. It allows us to test and refine design choices and software early on. We virtually run through countless scenarios to see how the robot behaves. This allows us to identify potential problems before physical prototypes are even involved. That speeds up development and reduces costs. And above all, it minimizes risks. After all, we know at an early stage whether the system does what it’s supposed to do. Moreover, because the software has already been largely tested in simulation, it’s ready as soon as the first physical prototype is available.

Our strength: the system as a whole

What makes navigation under challenging conditions so difficult is that there is no off-the-shelf solution you can simply buy. It requires the right combination of sensors, robust positioning, and navigation software that can handle the real world. By treating all these components as a single, integrated system, we build ground robots that keep moving where standard solutions fail—whether in the barn, out in the fields, or in the concrete plant.

We have incorporated that expertise into our RIBOTIC Navigation platform, which provides positioning and navigation for ground robots both indoors and outdoors, and is suitable for a wide range of robot configurations.

Are you working on a mobile robot that needs to navigate challenging environments? We’d love to help you figure it out. Get in touch and discover what’s possible.