In the first of our AMR navigation blogs we’re looking at the role robust localisation plays in accurate AMR navigation. If you’re building an autonomous mobile robot, or AMR, you’ll no doubt have already started to think about how you can enable it to navigate accurately and correctly in the environment you plan for it to be used.
In this blog – the first in our series on AMR navigation and localisation – we’re discussing how you can provide accurate localisation data – that is, data that tells your AMR where it is – in a variety of different environments. We’ll also touch on how to create a platform that can transition between different environments while preserving that accuracy.
What environments matter in localisation?
When it comes to localisation, there are really only three environments that we care about:
- Open-sky environments
- Partially covered environments
- Completely covered environments
If you haven’t already guessed, the defining factor in these environments is the strength of GNSS signal you can get. GNSS continues to be the gold standard in localisation data because of its global coverage, ability to deliver differential corrections, and its accuracy. However, as GNSS coverage drops, you need to consider alternative methods of providing your AMR with localisation data. Let’s look at those three environments now.
Open sky environments
Examples:
- Anywhere outdoors where you can see lots of open sky around you.
To give your autonomous mobile robot global position data, you need a GNSS antenna to receive signal from the GNSS satellite constellations, and an inertial measurement unit (IMU) to provide additional data about orientation, pitch, roll, and velocity. Most people use an inertial navigation system, also known as an INS or GNSS/INS, for this.
Things you will need to consider as you build your robot include:
- Do you need quad-GNSS? Quad-GNSS gives you access to data from the four major GNSS constellations: GPS, Galileo, GLONASS and BeiDou. It increases the accuracy of your AMR by increasing the number of satellites it can get position updates from and the different positions they occupy in the sky (allowing for more accurate triangulation). Quad GNSS is also vital if you want your AMR to operate anywhere in the world.
- For maximum accuracy over time, you will need a way to receive differential corrections. Because of the shape and rotation of the earth, corrections are vital for keeping an AMR on track over longer distances.
- Do you need global position? There are other, more cost-effective methods of providing your AMR with position data if you don’t need global position data. Some brands of autonomous mower, for instance, use a perimeter wire to sense when they need to turn. You need to consider your use case, your environment, and the commercial viability of your platform when deciding what localisation methods to use.
Partially covered environments
Examples:
- Urban canyons
- Forests
An AMR operating in a partially covered environment can be suceptible to a reduction in the amount of accurate position updates it can receive. This is primarily due to a lower number of satellites in view. This can subsequently lead to a degradation in navigation performance. In urban environments specifically, the concentration of buildings with reflective surfaces can also create multipath errors. If you’ve ever attempted to track someone running the London Marathon using their smartwatch, only to find that they are apparently in the middle of the River Thames, then you may have experienced a multipath error.
In these environments, we advise using additional sensors to augment GNSS. Your IMU data does this already if you’re using a GNSS/INS; beyond that, there are two options to try:
You can look at integrating additional sensors such as wheelspeed sensors to reduce inertial drift (which can be a problem when you have no GNSS signal). Alternatively, you could use a LiDAR scanner and OxTS LIO (see the boxout for more information).
You can attempt to make better use of your satellite data. OxTS’ gx/ix tight-coupling algorithm, for instance, delves into the data your GNSS antenna sends to your navigation engine to provide accurate navigation data even when only one satellite is visible.
OxTS LiDAR Inertial Odometry (LIO)
OxTS LIO is a technique for improving position accuracy in poor GNSS environments. It analyses data from a LiDAR to estimate your AMR’s velocity and angular momentum; that data is then used to improve position accuracy. Below is an example of how OxTS LIO improves position accuracy in a multi-storey car park. The navigation data used to create the pointcloud in this example was supplemented with zero velocity update data from the LiDAR. This constrained drift when GNSS was unavailable leading to a much higher percentage of accurate navigation data which in turn led to a better final result.
OxTS LIO has been demonstrated to work in real time, and we are working hard to bring real-time LIO to all OxTS GNSS/INS models.
Completely covered environments
Examples:
- Indoors
- Underground
In completely covered environments, there’s no GNSS signal available at all, so you need to find another way to provide your autonomous mobile robot with localisation data. There is a variety of ways you can do this. Let’s start by looking at infrastructure-based solutions vs no-infrastructure solutions.
Infrastructure-based indoor localisation
If your localisation solution is infrastructure-based, it means you’ve installed equipment in the environment for AMR navigation. Some of those solutions are powered, others are not.
An example of powered infrastructure-based localisation is an ultra-wideband (UWB) solution such as OxTS Pozyx2GAD. These solutions have traditionally given the most repeatable results (meaning the position data is more reliable) , but also require power to run and need you to survey your environment before you can use them.
Examples of non-powered infrastructure-based localisation include using ArUco markers and a camera mounted on your AMR to navigate. They’re simpler to set up (all you need to do is place the ArUco markers in specific locations and survey those points), but are generally provide less repeatable data.
Non-infrastructure-based indoor localisation
Unsurprisingly, this is where the platform can navigate indoors using no external equipment. There are two main technologies to talk about here.
The first is SLAM (simultaneous localisation and mapping). SLAM uses sensors like LiDAR or cameras to build a map of the environment the robot is in; processing algorithms then enable the robot to localise itself within that map and use it for navigation.
SLAM is a well-established technology, and works well in smaller areas. Across larger areas, though, accuracy can degrade.
The second is OxTS LIO. As mentioned in the boxout, OxTS LIO can significantly improve accuracy in GNSS-denied environments, without any infrastructure.
So, we’ve covered different ways of localising your AMR in a variety of environments. But what about moving between environments – specifically between outdoor and indoor environments?
The holy grail: transitioning between environments
Being able to transition between indoor and outdoor spaces is a vital differentiator for autonomous mobile robot manufacturers. It allows you to create AMRs that can leave and enter their indoor storage by themselves, instead of a human having to move them. It makes it possible to have autonomous forklifts that can move between a storage yard and an indoor warehouse.
Some of the biggest technical challenges with transitioning between environments are to do with time measurements, and with navigation frames. Both highly technical topics, but to summarise:
- Usually, your AMR will get its time measurement from the GNSS signal. If it starts indoors with no signal, how does it measure time – and how does it synchronise with GNSS time when it goes outside?
- GNSS position information is in a global navigation frame – it’s an absolute position on the earth. Indoor positioning solutions are in a local navigation frame – a set of x,y,z coordinates mapped to the specific location of the AMR. How do you move between those environments?
Other challenges, and perhaps just as big, include those associated with error modelling/fusing data correctly and also ensuring each of the localisation methods either side of the transition are accurate in the first place.
Of course, you don’t have to solve those challenges. You could partner with an organisation that has already solved them. At OxTS we’ve demonstrated seamless transition between outdoor GNSS navigation to UWB, and back again. We’ve developed technology that enables AMRs to initialise indoors with no GNSS signal, and we’ve developed a range of hardware and software features designed to make building an AMR navigation solution as simple as possible.
Autonomous Robot Navigation Solution Brief
AMRs need a robust robot localisation solution; a tool that not only records the position and orientation of the robot, but also operates both indoors and outdoors.
This solution brief steps through the aspects we recommend our customers consider when deciding on their source of localisation for their autonomous mobile robots.
Read the solution brief to learn how the right robot localisation solution can help your AMR project, including the key questions you need to ask yourself before embarking on a project.
If you’d like to learn more about what we can currently do for AMR engineers, click here to view our application page. If you’ve got a specific project that you’d like to talk to us about, contact us to get in touch. We’re always keen to help.
