Automated guided vehicles have been around for some decades, finding a small place in warehouses and manufacture to move materials and feed production lines. But in recent years they have become widespread and could soon be ubiquitous, part of growing robotisation throughout industry, often replacing conveyors, hoists and manual trucks.

"In the 1990s there was a steady increase of 5% annually and then perhaps 10- 15% " says Petri Petäys, sales director at Finnish manufacturer Rocla, part of Mitsui Mitsubishi but specialising in AGVs of all kinds.

This trend might increase even more. He says acceptance is growing fast for various reasons including a new robot acceptance in general culture helped by developments like the Google Car and other motor manufacturers’ interest in driverless vehicles. Other factors include: a turn towards AGV manufacture by volume forklift manufacturers like Toyota, who while new competition are also popularising the concept; and the mass availability of once expensive components like laser scanners. The greater use of electronics in standard forklifts is also blurring the boundaries.

A world wide slow down in production and therefore investment since 2008 has perhaps held back what would otherwise have been a bigger shift, he says, but even so automated systems have been seen as an answer to rising labour costs and pressures for greater productivity.

According to Nick Bellwood, European sales director for US firm JBT, some think the market could top $2bn by 2020. "AGVs do not stop or have lunch breaks and they work steadily around the clock" says Neil Smith, a UK engineer for Swedish maker AGVE Group "That makes them very cost effective for three-shift 24hours operations especially." The firm works alone and with partners like international logistics company Swisslog.

The application of AGVs is already multifarious. They have been widely adopted by the automotive industry for example where they fit with robot manufacture, both to supply just-in-time production components, and even to carry the robot arms. Many other production processes use them, both for materials and as moving assembly platforms. The lean production notions of modern factories are served well by their steady operations.

They also frequently replace the production lines themselves, with units moving from work station to station, instead of a conveyor or chain system moving the assemblies along.

"That means processes can be easily reconfigured," says Gary Koff at Savant Automation in the US" unlike a chain driven line." It also means a holdup at one station does not stop the whole line not breakdowns; an AGV production sequence can also have single moving units removed for maintenance or repair while the rest continue.

A major growth area is in storage and warehousing and makers such as Belgium’s Egemin cite particularly the food and beverage sectors, which are growing fast. Italian maker Elletric80 also does a lot in this sector, along with Rocla and major producer JBT from the States.

The advantages are particularly good for clean room areas or cold storage but also for high racks, narrow aisle storage and unlit racks.

"It is vehicles for these sectors areas which are the biggest growth," says Jayesh Mehta at US firm Transbotics "usually carrying between 500kg to 2000kg. That is the sweet spot at present."

Servicing these sectors means a variety of types of vehicle are produced by most of the makers. The majority are some variation on the fork lift configuration that is considered the most versatile, working mostly with pallets. Looking on many AGV websites, there is a range of sizes and shapes offered for handling different racks and storage units, and the loading points they service.

There are also tuggers, used to pull along one or more loaded trailers, making them the most productive, and there are assorted platforms with rollers, chains or small conveyors on them, which can move block loads and boxes etc from racks and stacks, or can connect onto production conveyors to transfer loads.

AGVs are even used for some people carrying functions; JBT has supplied mobile platforms for tourists visiting the famous First World War battlefield at Verdun in France, says Nick Bellwood, European sales director.

At the smaller end of the market are automated guided carts, much smaller units for carrying all sorts of materials. These are often found in hospitals where they lift and transport linen trolleys, mobile medicine dispensaries. This smaller units often come as a basic low profile motor unit, which can slot underneath a range of trolleys and other passive carrying units, lift up to connect a pin, and then move them as required, dropping them off for unloading or loading and moving on to pick up another.

At the other end of the scale are bigger specialised units for handling rolls and slabs in industries like paper and plastics, with grabbers and clamps to hold cylindrical and other shapes. Some can rotate and orientate rolls during transit, ready to fit onto a production or printing press.

Most AGV companies also supply very large, usually customised, units too, for industries like steel and other metal foundries. This can be quite a demanding sector and one or two firms have more or less specialised in it, like Finland’s Solving company.

Pekka Joensuu, AGV systems manager, says: "We do platforms from about 5t up to 120t capacity for steel, big production like heavy excavators, paper mills and aircraft." The firm’s production line platforms carry the Airbus A340 wide bodied plane sections in the assembly areas of the factory for example.

These bigger units are huge, sometimes multimillion investments and often have a much longer lifespan than the smaller AGVs. "Typically they might see a mid-life upgrade and renovation at around ten years," says Joensuu.

Jayesh Mehta at Transbotics concurs with him. The AGV design and supply company covers the range of machines from small to large including such largescale custom units and often goes in to upgrade the control systems later. "Sometimes they are still using Windows 95," he says.

His comment reflects the complexity of the industry and what it supplies. Firstly, there are the vehicles themselves, often customised and tailored versions of basic carrier units, though in many cases they are specifically designed, using a range of components. In some cases these will be supplied by subcontractors with the AGV company itself putting them together as required. "Transbotics works in this way,"

says Mehta explaining that "we are an engineering and design company really". For very large items they partner up with a supplier of usually manual equipment and jointly produce an automated unit.

That involves a second level of design with the addition and integration of an autonomous navigation system for the vehicles with its drive and transmission. These have developed from early wire guided systems to modern laser systems, both using reflectors and increasingly the so-called "natural navigation" using 3d laser scanned imaging of fixed objects and building elements.

Savant’s Koff has seen developments from early days. He started in the industry with one of its originators, Arthur MacBarrat, who developed a wire following tug unit in the 1940s for grocery order and dispatch. "They even put the wire in the air at first but it kept getting snagged so it was put into the floor."

This developed into various systems such as tape on the floor, and also buried magnets. Later in the 1990s with the advent of spinning lasers, a new system developed based on laser lines from strategically positioned reflectors, mounted on walls, columns and other building elements. The machine computes its location from triangulation between three reflector signals to a high degree of accuracy.

"The reflectors are fixed and then carefully surveyed during installation," explains Petri Petäys at Rocla.

Laser location is now the most commonly used system for AGVs but it has some drawbacks including the need to maintain line of sight for the units, and the possibility that reflectors can get knocked off or moved out of alignment.

Koff’s Savant offers a different system using magnets combined with inertial navigation. The latter relies on a gyroscopic system to give the vehicle a dead reckoning location. "It will check its position as it passes over a floor mounted magnet and correct any error in the line" he says. Like reflectors the magnets are surveyed on installation.

"The units then have routes programmed into them with a simple CAD drawing type of representation," he says "which can be fairly easily changed if you wish to reconfigure the layout."

Savant makes the vehicles and their controls, but as the navigation becomes more complex a number of companies have been specialising the vehicle control only, guidance as a component to the AGV makers themselves such as Transbotics. The best established of these companies is Kollmorgen, a major European motion technology company with an AGV division located in Sweden.

Originally it is was part of a group making AGVs but decided to focus on the controls, says Mehta, from Transbotics which split away from Kollmorgen over a decade ago. "Now we don’t make the AGVs but different users fit our control systems," says Peter Bladh, product development manager. "We work with them on the design of vehicles when it comes to fitting the controls. Kollmorgen also supplies vehicle control software although many of the makers then modify that, or develop their own.

Kollmorgen’s systems are primarily based on laser reflectors but it offers various methods and hybrid systems to, that allow AGVs to switch between different control methods as appropriate. Its latest product is a so-called Natural Navigation system, launched early this year after years of research. This takes advantage of the safety lasers which are fitted on to nearly all AGVs as standard. Safety lasers are sometimes known as "bumpers" because they replace physical bumpers mounted on early machines, which would stop a vehicle if it hit an obstruction or more importantly, apply the brakes if a human was in its path.

These lasers have developed on a parallel path to the control systems and are supplied by companies like SICK and the UK’s Keyence. They measure time-of-flight for the laser signal to scan a field ahead of the machine and detect any obstacles. "Mostly they use multiple fields," says John Slater at SICK "and first slow and then stop the AGV by applying the brakes and cutting power to the motors." The most sophisticated versions can alter the size and orientation of the fields as they slow incrementally or go around corners, scanning usually close to the ground.

The lasers have been important for increasing the operating speeds of AGVs, which can travel at 2 to 3m/sec without jeopardising safety; most AGV makers are keen to emphasise the relative safety of their vehicles compared to manual forklifts which cause numerous industrial accidents annually, including fatalities.

But these scanners are now being coopted by the likes of Kollmorgen taking advantage of the time-of-flight data they produce. They can provide an image similar to the 3-d point cloud images produced by survey laser scanners. And while not to the same standards, the data is sufficient to feed to the vehicle control module where it produces a basic 3-d map of the surroundings.

"By using that data, you can use elements of building such as corners, machine edges and columns as fixed orientation points," says Bladh at Kollmorgen explaining "the machine can build its orientation and set a path using those." He means it is unnecessary to fix and then survey laser reflectors, or install other guidance infrastructure like magnets or buried wires.

"Setting up is easier and flexibility increased for when production is restructured and new vehicle paths are needed."

The new system is just winning acceptance and some makers are still cautious, believing the accuracy is not yet to the millimetre levels needed to position an AGV onto a conveyor belt or to plug it into a power charging point. This is something that all the electric battery machines need to do occasionally.

Another caution cited is that natural systems – other makers also offer it then – are sensitive to changes around them, having difficulties if objects are left in a vehicle path, such as a stack of empty pallets. Making too many changes, they could become disorientated.

These objections do not apply in areas like hospitals where the long corridors are consistent and unchanging, and natural system have already caught on much more there.

"You can also have difficulties in wider spaces where the laser range is not effective," says Jouni Sievilä a control software engineer at the small Finnish firm of Navitec, which has a long history making self-moving systems for mining equipment, trucks, loaders and drills. It has recently entered the AGV sector.

"Mines are easier because there are rock walls all around but factories might have fixed points 30m away."

The answer he suggests is hybrid methods, using natural navigation software but supplementing this with reflectors in awkward or difficult areas.

New aspects to this technology include dynamic map building in the onboard computer model. "At present the vehicle is driven through its route and memorises it statically," he explains "but the next thing is to change the route according to changes in the environment or obstacles it finds."

There are some makers already offering such systems. One is Canada’s Clearpth that offers a modularised system for the lower and mid-size end of the market, based around its autonomous vehicle christened Otto 1500, the number referring to its 1500kg maximum load. It has various load carrying configurations for mostly pallet carrying.

The OTTO, and a recent small cart version introduced, uses so-called simultaneous localisation and mapping, or SLAM. According to Simon Dexter, sales director, it can build an internal map from a "walkaround" to locate natural features and will then find a path from A to B as required. This path might vary from time to time and will include a capacity to go around obstacles if they remain too long in one place.

Eventually, there may also be systems like this using visual guidance with cameras, and emerging shape recognition software, though it is still a future option at present.

While this technology is developing, AGV control involves a third level. This is the fleet control itself. As well as navigating, the units have to accept orders, find a loading or starting point and move to and end point, perhaps connecting to another machine with a "handshake" signal, or recognising an empty cart or contained to pick up.

That can be very complicated and one of the constantly evolving areas of AGV provision is the "traffic control". Most AGVs are linked to a central command computer, usually by a local radio system. This control has to accept orders for movements, and then work out how to dispatch machines.

It needs to track where units are currently, and their status, full or empty, returning or on their way to an assignments. It also requires to monitor their condition, particularly battery charge levels, to ensure they are capable of tasks.

"Then the system needs to find the most suitable unit for a task", says Koff at Savant. The nearest one, of the right capacity and without a task has to go. It has to set priorities for different units that might travel the same path.

The central system also has to deal with possible traffic jams, especially in complex factory layouts, which might involve a variety of routes. Jams can be caused by obstructions or by two vehicles attempting to enter the same space – usually systems use a cell-system like a railway block system where only one unit can enter.

Another aspect of these control systems is integration with warehouse or production computer inventory and component tracking systems, usually those owned and operated by the client. In future this may mean AGVs picking up and monitoring more details of their loads and where they have gone.

The algorithms for this control can be very complex and especially as the number of vehicles in a fleet climbs higher. Computer power needed to track and assess options for 100 units can be significant, particularly for traffic optimisation which goes beyond simple rules.

In fact most of the makers say that around 100 is the limit at present. Virtually all the AGV makers emphasise that such control systems are an integral part of their offering. It means their product is not simply a they offer is not simply a machine, but a system. It will be devised and designed in conjunction with the manufacturer and his overall production or storage layouts "and sometimes we will be better at knowing his logistics than he is himself," says one. Machines, though built around a basic design, will be customised and tailored for those logistical needs. One new development might be a new level of autonomy for units, where instead of AGVs being assigned and routed by a central control, they will interact machine to machine in order to work out routes and priorities.

Such a system is being explored at present by Dutch maker VDL Steelweld, which has specialised in autonomous container movers, one of only two firms in the sector along with Terex, though several others are eyeing this market. VDL itself has supplied Rotterdam docks and Singapore’s huge container terminals, with over 125 units to date.

Dockside AGVs are a highly specialist area, involving larger units than usual and with additional design requirements. Units are 18m long and accept up to a 78t payload.

"You have to consider aspects like asymmetrical loading when two smaller containers are carried with very different contents perhaps, tyre pressures for the big wheels used, and the outdoor environment," says Karel Smits.

There are longer distances needed for harbour work, at least 1-2km rather than a maximum few hundred metres in a factory or warehouse. All that precludes certain types of guidance system; natural navigation has problems in an exterior world of changing light conditions and bright high contrasts one day, and rain the next. "Even the safety scanners need close attention to prevent them being obscured in say, snow, or rain," says Smits.

"For this work we did explore GPS systems, which have been used in agriculture; you need an RTK corrected signal for sufficient accuracy and doing that cost effectively is a first challenge. But in a cluttered or urban environment it does not work well.

Consequently, VDL units use a gyro system combined with buried markers, mostly RFID transponders, or magnets, which are used to correct any path errors in the inertial control.

"We developed this from a system used successfully for bus routes in several locations, Eindhoven, Istanbul and in South Korea."

"The technology is accelerating now, particularly with Google cars and the like, and sensors and equipment is much easier to get that you once had to make yourself." VDL is running a test vehicle in Rotterdam which embeds the latest thinking he says to try and overcome traffic holdups, "which can occur even where you don’t expect them". The port operator has to constantly watch the system screens to be sure a vehicle has not halted at an obstacle, holding up multiple units behind it.

Units are currenlty giving instructions from a central control, but the new unit has an autonomous capability, he says taking something from the flying controls being used by drones to avoid clashing. It should be able to handle obstacles better and find a way around them.

VDL’s next moves are outwards into other large external production areas and it is working on a project with a large chemical producer in Rotterdam.

He is also convinced that the technology will find its way into public transport more and more, in the Netherlands at least. And given the drone system developments, you could say the sky is the limit.

Finally, there is a third component, the control and command system to manage the machines, their routeing and interactiThis traffic control is usually a central computerised system with some fairly advanced and complex algorithms, communicating with the individual vehicles by wireless in most cases. It will often integrate with a client’s management systems, either a warehouse tracking and ordering software, or a production line control system.

Such a command system can be quite complex and particularly when fleets of automated vehicles get larger. Often a client will have perhaps five or ten units but bigger facilities have dozens and as many as 250 can be operating.

A key point made by many of the suppliers is that they are not selling machines but systems for materials and production; although they will have basic AGV units built around a modular principle, that is by no means the whole of it. Both the navigation systems and the traffic controls add layers of complexity.