A magnetic track is laid onto or into the floor, with the line of the magnetic track following the centreline of the vehicle’s route through the facility.
On the vehicle, one or more magnetic sensors track the position of the magnetic strip on the floor. As the vehicle moves and the position on the magnetic strip changes, changes are made to the vehicle steering and drivetrain, to allow it to follow the magnetic strip.
In many cases, the route can contain junctions and branches to allow for loading and unloading or charging without blocking the main line.
A magnetic track will need to be laid everywhere a vehicle is required to transport something or perform a task.
A magnetic track will often not suffice for localisation and control. Without assistance, a vehicle on even the most simple of circuits will not be able to determine exactly where it is located. To provide this functionality, additional means of providing commands and location are used.
In the earliest systems, coloured labels, like large multi-coloured barcodes were used. Contemporary systems use groups of magnets, RFID tags, standard barcodes or QR codes. Upon coming across one of these items, the vehicle will “read” the item and determine both it’s location and the next step of it’s program to execute.
Generally speaking, a magnetically-guided AGV will follow a set program, continuously driving and stopping around a circuit and responding to guide tags as they are encountered. Program branching can be triggered by a guide tag, vehicle status (such as battery level or a fault) or through an external means such as a capture box or other machine interface.
Magnetic guidance, whilst an older technology provides a very reliable means of AGV guidance. Some systems allow the lifting a re-laying of the magnetic strip and guide tags to re-route the system, when required. This is particularly popular in fast-changing production environments such as the automotive sector.
This ease of lifting the strip is also a downside in that long-term maintenance is higher due to the increased risk of magnetic strip being inadvertently moved (for instance by manually driven lift trucks).
Laser guidance is by far the most popular method used today. A laser-guided AGV is fitted with a LIDAR. Reflectors are placed around the proposed working area and their locations are provided to the AGV. Laser light emitted from the LIDAR bounces off of the environment around the AGV and returns to the LIDAR. The time this light has taken to return (called “Time of Flight) allows measurement of the distance. Light that strikes a reflector is polarised by the reflector – this enables the LIDAR to “see” only the light returned from the reflector. The concept is close to that of triangulation – where a bearing is taken on three or more landmarks. The intersection of these bearings indicates the current position of the vehicle.
In a typical system, a minimum of three reflectors are required for operation, although movement accuracy may suffer. In practice, 5 or more reflectors are preferred. Reflector positions within a working area have their positions pre-determined before installation, as the overall pattern of placement is key to an AGV being able to locate itself within a map.
The transport circuit positions will be defined virtually, in relation to the reflector positions. This is important to understand, since adjusting the position of the reflectors could re-position the AGV view of the location of the transport circuit.
LIDAR technology has become increasingly cost-effective over the last 10 years. Individual reflectors (either tape, cylindrical or triangular) are also relatively cheap to acquire. The challenge for laser guidance is in the expertise in determining reflector positions and getting them installed. In larger installations, this can be a challenge although depending on the system being installed, there is some brilliant software available to ease this process. Whilst the technology is more advanced, a laser guidance system can be less flexible than magnetic guidance. Reflectors are less easily moved and re-installed as they are often placed at height and need physically fixing to a surface. The reflector map within the AGV system will also need updating/re-calculating, which currently requires the services of a specialist.
This method of guidance also utilises a LIDAR, but does not require the use of reflectors. Instead, where traditional laser guidance “looks” for a set of specific targets, the entire return is used in the localisation calculation. This means that instead of 3 or 5 “bearings”, there can be many thousands instead.
During installation, an AGV is driven around the proposed working area and the environment is “mapped” by the LIDAR. This map is processed, to remove non-permanent items (such as stacked pallets, passing equipment) and a transport circuit added. Once the map is back within the AGV, the vehicle can process the return from the LIDAR and compare it to the map and odometry calculations to get a precise estimate of the actual position.
Natural Guidance has become more prevalent over the last 5 years and will continue to do so. This has been driven in part by the increased availability of more powerful embedded computing and advances taken from research into self-driving vehicles in the automotive sector. Due to the nature of the LIDAR measurement and often deep fusion with the vehicle odometry, Natural Guidance can be quite tolerant of changes to the environment, without needing the map to be updated. It’s important to note that the result of the LIDAR/map/odometry calculation is a probability of a position. A high value means that the vehicle has high confidence in it’s position, with lower values meaning less confidence. This means that if a wall or column is physically removed, then all that happens is the position “confidence” reduces – this wouldn’t be possible if we removed reflectors or magnetic tape entirely.
Some Natural Guidance systems can update their map as the environment changes. For instance, if a wall or column is removed physically and remains missing for a pre-determined length of time, the AGV can update the map to include this change. This is generally called Simultaneous Location and Mapping, or SLAM for short.
Natural Guidance only works where the environment can be “seen” directly by the vehicle. In areas where there are tall, temporary structures (such as stacked pallets in a staging lane or block storage) the LIDAR may not be able to see most of the building that it expects. This problem gets worse the lower-down a LIDAR scanner is placed as now most temporary objects (such as pallets, people and other vehicles) are more numerous.
Some of the very latest Natural Guidance systems have replaced the LIDAR with a set of cameras. This type of system utilises vision processing to provide a 3D representation of the environment around the vehicle. Camera modules tend to be less expensive than LIDAR modules, so more of them can be used. This enables cameras to be viewing multiple planes outside of horizontal (such as the ceiling), which gets around the potential blinding of LIDAR systems when surrounded by temporary objects.
In the right application, with the right vehicle, Natural Guidance can be amongst the least impactful and quick to install. This is due to there being no physical infrastructure (such as reflectors) required to be installed and a shorter mapping process.
Navigation for an automated vehicle can be a complex topic. There are two methods that are currently widely available:
Fixed Transport Circuit
This is the “traditional” way of having an AGV navigate around a working area. Either physical (in the case of magnetic track) or virtual drive paths are defined. An AGV ensures that it stays on the defined drive path as it drives around the transport circuit. Definition of a route between two points can be defined up-front or in some systems is calculated on-the-fly according to current traffic conditions.
Vehicles that follow a pre-defined transport circuit are referred to as an AGV. Their movements paths are predictable, meaning that they can be marked out, making working interruptions less likely. When an object is in the path of an AGV, it is also unlikely that an AGV will be able to navigate around it – instead it will have to wait until the obstruction is cleared.
This type of vehicle is commonly referred to as an Autonomous Mobile Robot (AMR) and is considered (usually in marketing terms) to be much more intelligent than an AGV. The reality is that an AGV and AMR are very similar, differing only in the operating software and operating parameters. An AMR is generally able to traverse a working area having defined their own transport path between two points. This can include object avoidance, provided there is enough space to define a replacement route. This can make implementation much easier but can bring some unpredictable behaviour operationally – a fast-changing environment can confuse an AMR trying to find a route to its destination.
No-go, speed restriction and other controlled areas can be defined for an AMR in much the same way as segments of a traditional AGV transport circuit. This means that control can be exerted over the driving behaviours of a typical AMR system.
What should I specify or choose?
The simple (and unhelpful) answer is that the choice of guidance or navigation should really depend on the application under consideration. This means talking with the experts directly, asking them to evaluate your requested application and make a recommendation. Once this recommendation is made, challenge the technology choice offered and ask:
- Why the offered/recommended technology?
- Are there any alternatives?
- If alternatives are available, what’s the difference with the recommended option, with respect to the application under consideration?
A great supplier that wants to work with you will enthusiastically engage with this process. At RM Group, we’ve many years experience in automation – not just designing, manufacturing and integrating but also in compliance& aftercare.