This article is the fifth of a series dedicated to robotics, which will be completed within the next month.
We define robots as intelligent machines possessing a degree of autonomy, with three characteristic features: sensors to understand the robot’s environment; on-board processors to analyze the sensor data and take decisions; actuators (motors, tools and articulated arms) to act on the outside world. Robots have already been present in industrial uses for several decades and today are becoming operational in the service sectors. But the challenges and stakes are quite different. To better understand their new uses, let us dwell for a moment on the connotation of the somewhat vague terms “service sectors” representing, as they do, more than 70% of GDP for countries with developed economies.
In economics, a service consists of making available either a technical or intellectual skill for either a private person or for a business enterprise. Often, we oppose services and industries, insisting on two points: in a service, work accrues directly to the end-user (viz., an enterprise or a private individual) whereas, during the service process, there is no transformation of matter.
This latter point is controversial: a restaurant owner, a hairdresser, a surgeon professionally transform matter. A finer analysis would, however, distinguish the industrial world on one hand (i.e., a closed sphere where operatives and machines transform matter in mass-production processes and, on the other, services seen as sets of activities that occur in various spaces (at home, in the street, in enterprise). The impact relates to humans (medical practitioners, hairdressers…), or to their close environment (household chores, safety monitoring, surveillance, replacing or assisting humans for certain specialist activities (banking counseling), or simply in facilitating life-styles in terms of travel arrangements or in making goods available (trade, logistics).
The variety of tasks that robots can handle is similar to that of the human actors. Some services mobilize public organizations while other involve entire enterprises, and yet others are an integral part of the informal economy, for instance ‘baby-sitting. Some services address private individuals in particular, others aim at organizations, notably enterprises. But underlying the large range of services possible, we find a simple idea: the main feature that characterizes a service is its interaction with the world of human beings, as opposed to industry which interacts directly with matter.
In this context, Olivier Fallou and Robert Millet (at a seminar organized in 2012 by the French Interministerial Pole for Foresight and anticipation of economic changes), introduced the topic of service robotics by insisting on the fact that it represents “the capacity to operate in an environment initially designed for men, interacting with an environment designed for men and interacting with humans.” When we say “service,” we do so in reference to a human world. Hence certain specific constraints, such as, for example “guaranteed (fail safe) operation enabling operation with the presence of a possibly large audience, and in this case, composed of non-professionals.”
What does the robotic service sector represent today? It is an emerging market - following in the wake of the success of industrial robotics – which, today, is a rapidly expanding sector. According to the International Federation of Robotics (IFR), business could represent 100 billion euros in 2018 and 200 billion in 2023, compared with 25 billion in 2015. The robot assemblers are almost all SMEs and start-ups established by scientists coming from research laboratories, especially in the area of personal aid robots. Some major companies active in developing and assembling industrial robots are monitoring certain applications of service robots, either directly or via laboratory start-ups (such as Kuka in Germany or Fanuc in Japan).
The wide variety of actors also is also reflected in the range of products. There are, for example, humanoid robots like Baxter (whose head is a screen), who can carry out simple tasks he is taught to execute and can detect the presence of human and thereby salute him/her. Humanoid robots are not alone in this arena. The International Federation of Robotics has noted the multitude of shapes and structures and, likewise, the wide range of application areas. Service robots are not necessarily mobile nor self-standing. “In some cases, they consist of a mobile platform on which one or several arms are attached and controlled in the same mode as the arms of industrial robot. Furthermore, contrary to their industrial counterparts, service robots do not have to be fully automatic or autonomous. In many cases these machines may even assist a human user or be tele-operated.”
In a widely recognized work of reference publication, in The Quarterly Journal of Economics, Nov. 2003, authors David Autor, Frank Levy and Richard Murnane asserted that the fundamental criterion for replacing human labor by machines, is both the routine nature of the tasks undertaken and their inherent simplicity; under these conditions the distinction between service and industrial robots is barely significant. And the observed wave of robotization would not make the distinction either.
The services that one can robotize today (or potentially tomorrow or later) are characterized by repetitive tasks with limited choices, because, if a robot can be designated as having a degree of autonomous movement associated with the possibility to make decisions (on all or part of the elementary tasks set), it nonetheless remains inadapted faced with complex actions that require taking initiatives. That is, except in those cases where a human is in charge to accompany the work, as is the case for drones or robot assisted surgery.
Numerous activities can be robotized, among which logistics and transport, industrial cleansing, surveillance and maintenance (sites, public access areas, infrastructures), on construction/demolition sites, in medical assistance, domestic assistance, aid to personal independence …
Let us try now to better understand the underlying mechanisms. A first distinction is to identify two main categories: professional robotics and personal robotics.
Alongside military robots (Cf. this series #2, Can military robots embody an ethical stance?) and robot surgeons (Cf. this series #3, Your next favorite surgeon might be a robot), the first category includes robots that perform tasks in a public access area or in an enterprise. Among these, industrial cleansing and cleaning of public access areas, delivery robots for offices or hospital services. Intervention robots (on-site terrain robots, fire-fighting robots, demolition robots…) also belong to this category, characterized by the presence of a specialist robot operator, who responsibility it is to start and stop the robot and also to monitor work in progress.
The second category includes “general public” robots, such as the robot vacuum cleaner already present is some of our homes, but likewise robotic machines designed to meet the special needs of handicapped or fragile persons: automated wheelchairs, aids to better mobility, pet companions or animal exercise units. The International Federation of Robotics defines this category by insisting on dual fact that these are “non-commercial tasks” and also that they can be operated by lay persons, without any special training.
The professional robot market is dominated by defence robots (Cf. the second article in this series) and intervention robots. Both areas of application are characterized by the robots being used in dangerous environments. Intervention missions can be assigned to robots, for example, to access installations or equipment during a fire or to premises potentially contaminated by chemicals or by radioactivity, to observe, measure, sample and also carry out various manoeuvers.
The Intra Group, which owns the first global ensemble of intervention robots, was created in 1988 by the three major French nuclear industry players: EDF (the electricity utility), the CEA (Atomic Energy Agency) and Areva – then known as Cogema); the Group equipped itself with a fleet of robotized, semi-autonomous mobile units developed by the CEA and certain industrial partners such as Cybernetix. Among these units is a noteworthy robot consisting of two articulated bodies that enable it to climb and descend stairways and even negotiate obstacles 40 cm high! To achieve the latter, the robot is capable of raising either its fore or rear part, as Jean-Marc Alexandre, CEA research engineer explains (in French). The possibility for such movement is obtained by displacing the manipulator arm, thus affecting the centre of gravity of the two-part machine. To resist the effects of destructive radiation, the robot is fitted with ‘radiation-hardened’ electronic circuits that can withstand radiation levels up to 100 times higher than traditional electronics.
Intervention robots look more like military vehicles than robots. Guardian, TerraMec and Thermite, for example, constitute a team of robotic firefighters sold by the American company Howe and Howe Technologies. The Guardian is an articulated arm used to move debris; TerraMec is a mini-bulldozer capable of clearing a route to a site that needs dowsing and the Thermite is a fire-fighting unit that can deliver water at 2 270 litres/minute. All three units here resemble small tractors. In yet another application area, we find submarine robots - such as those developed by the European Centre for Submarine Technologies (with 32 research scientists from five EU countries: Portugal, Germany, Spain, Italy and France) – that can be used against oil spills, to localize oil surface slick oil. Having a human appearance can have its advantages, notably in environments specific to human presence. We read that in April 2014, the US Navy tested a humanoid firefighter called SAFFiR (Shipboard Autonomous Firefighting Robot), designed to fight ship fires on the open seas. The robot was developed by the research teams at the Virginia Tech, UCLA and the University of Pennsylvania. As Franck Latxague, founder of the web-site Humanoides.fr says, “these robots must be capable of carrying out various tasks such as, for example, keeping their balance when the ships rolls and pitches, opening fire hydrants, picking up objects, aiming the hose at the heat source and extinguishing the flames. They will be fitted with vision shape identifier sensors to identify humans through smoke screens. They can memorise the ship’s layout and move autonomously through its cursives and decks.”
In comparison with these urgent, emergency situations, the daily life of demolition robots can appear far more relaxed but, nonetheless, demolition sites are also dangerous for human operatives, with the risks of structural collapse and thick dust projections. Demolition robots offer high performance in this environment, for tasks that do not call for any special degree of accuracy. The Swedish group Husqvarna has developed a family of remote controlled demolition robots – highly powerful, manoeuvrable, stabile machines with long reach.
Beside their capacity to undertake dangerous missions in extreme, hostile environments, robots can prove specially adapted to routine activities, notably fastidious, tiresome tasks.
Let us analyse the case of Amazon. The e-trade giant has often been criticized over the work conditions in their gigantic warehouses, and had difficulties managing absenteeism, turnover and even pilfering by employees of items in store. This is where robots can prove useful: they represent a silent, uncomplaining, flexible and obedient workforce, who would never be tempted to steal from the stocks. Amazon in 2012 announced the acquisition of Kiva Systems, a company specialist in robotized warehouse, for 585 million euros. The objective of Amazon is to gradually replace warehouse labor forces by robots, integrated in a data processing system that calls for complete chain automation, from stock management to packaging and mail expedition preparation. The challenge, apparently, consists less of cutting back on costs than gaining in productivity. Robotizing Amazon’s warehouses will be notably decisive in terms of targeting handling speed and system reactivity, identified by author Brad Stone in “The Everything store…,” as the key features of the Seattle company to gain a competitive edge in Internet trade.
Logistics represents a growth market for service robots. The areas of application are not just limited to warehouses but include offices and hospitals, where small automated electric chariots can already be seen in use routing medicinal drugs to meet service/ward demands. In the warehouse environment, robots are mobile (forklift) platforms similar to those used back in the 1950s except that the modern versions are totally automated and can achieve a far larger range of tasks, therefore making human intervention superfluous.
In addition to these kinds of machine, there are also alternate systems, such as the one designed by Balyo, a company set up in 2005. Raul Bravo, an engineer from Ecole Polytechnique and co-founder of Balyo, explains in the French ICT magazine 01.net that, if the objective is still to automate dispatching of goods and items inside a warehouse, the solution he proposes is very different: a simple box called a MoveBox which enables a standard fork-lift or elevator truck to become autonomous. There is no need for guidance infrastructures on the warehouse floor. The MoveBox uses its own on-board 3D imaging analysis to check where it is. “We are the only suppliers in the world offering such a system without any floor infrastructures,” adds Raul Bravo.
Let us stay for a moment in the warehouse environment. Among the routine tasks that robots can carry out equally well if not better than human operatives, is to ensure monitoring functions. Thus, the e-vigilante robot developed by Eos Innovation can be programmed to do the rounds or beats, in the dark, using its on-board sensors to detect the smallest suspicious signal and issue a warning to a control officer who, depending on the circumstances, can then double check, clear the situation or take appropriate action. To do so, the officer takes over the robot controls and uses the on-board camera and loudspeakers. A procedure like this avoids putting a human supervisor’s life at stake, and improves the overall system efficiency. The aim is to reduce surveillance/monitoring costs while optimizing safety measures in sites under supervision. A robot has an added value thanks to its on-board mobile cameras, given that criminals otherwise make use of blind spots in the fixed camera fields. E-vigilante can move at an average 4-6 km/h, with a top speed of 10 km/h if necessary. The robot is small (35 cm high), but if someone tries to catch it, it will set off a strident alarm signal and blinding flashlights. The economic model for Eos Innovation is based on a robot hire-maintenance contract. Clients pay 2,500 euros/month to rent a basic model.
Apart from surveillance functions, certain restaurant tasks (particularly the tedious jobs) are now being robotized. Consequently, it does not come as a surprise when we find them in fast-food outlets, on an industrial chain model. A Californian company, Momentum Machines proposes a robotized hamburger confection line capable of cutting the tomatoes, cooking the meat and adding a sauce. This robot chain can turn out 360 burgers an hour! In China, a lot has been written lately about the Robot Restaurant in Harbin where some 20 robots serve at table and cook certain preparations such as the baoze or noodles. Restaurants in China, as elsewhere, represent a sector with acute labour shortage and recruitment difficulties.
Other unskilled tasks are currently being ‘automated’, for instance in gardening. Automatic lawnmowers, such as the Robomov model have been on sale for several years now. The majority of robot lawnmowers do not really possess an automated guidance system per se. Tracking is random; the surface to be mown is bounded by a string, which alone brings operational constraints and low efficiency. However, a recent innovation could change matters, especially for large surfaces (such as golf courses): the Nav on time system comes with satellite guidance, i.e., high precision positioning, a smart mower-path with user adjustable parameters and no need to mark out the terrain by string.
Over and above gardening possibilities, agricultural sectors in general are now beginning to use robots, occasionally for complex jobs. An example here is the so-called Vin robot which can prune vines, a tedious job for which the wine-growers find it difficult to hire competent labour forces. We are only at the beginning here, as notes Bruno Bonnel, President of the French robotics trade-union Syrobo, asserting that “agricultural robots are currently being developed for the tasks of weeding gardens or milking cows.”
We now address the most sensitive – also the most interesting - area for robotic potential and service robots, when they are designed to interact with humans. We can ignore the anecdotic uses in PR operations, where look-alike robots welcome visitors to public space events. This is understandably where we find the robot “stars”, such as the vacuum cleaner Roomba or the humanoid Nao. We are also dealing here with a different economic model, split as it is between mass-production machines and tailor-made niche models.
Service robotics for household chores is a market estimated, according to the International Federation of Robotics, “in year 2012, three million domestic service robots were sold, up by 20% on the 2011 figures.” If we can now quote the forecast for the period 2013-2016 – we read “Domestic robot sales could reach 15 million units, worth some 5.6 billion US dollars. The majority are toys, now at affordable prices. We can also look forward to seeing three million education or research oriented robots. Robots that provide assistance to persons with impaired mobility only represent 6,400 units sold, but this market segment will certainly increase considerably over the coming two decades.”
Europe is noted as being very efficient in professional service robotics (logistics, construction, assemble and demolition mainly), whereas leadership in domestic robotics is shared among Japan, the US and South Korea, the Prime Minister of which went on record as predicting (in 2006) that in Korea “we want to see a robot installed in every household by 2010.”
Japan buys 1/5 total world robot output every day. In France, 190 000 Roombas were sold in 2012, 21% up on the 2011 sales figures. Let us not be mistaken here, if Roomba has proven so attractive it is also because the performance price ratio has become acceptable (approx. 400 euros). Ratios like this play a very important role in market development. A robot lawnmower with a price tag of 5,000 euros will not have any clients, but with entry prices now at 950 euros, they have moved out of the luxury gadget category.
Most domestic robots, such as lawnmowers or vacuum cleaners, are in fact simply improved models of existing domestic appliances. If we seek novelty, we must turn to companion or personal assistance robots.
58 cm high Nao, one of the best-known robots today, is used to offer company for persons suffering from loss of autonomy, or autistic children or persons in hospitals. Several thousand Nao, have been sold by Aldebaran since 2008 and Nao’s successor Pepper probably will have the same commercial success. Competition is now building up with Honda’s humanoid Asimo.
Inasmuch as they are akin to robot animal pets such as Genibo proposed by the Korean company Dasatech, or other deskpets, priced at only several tens of euros apiece, humanoid companion robots offer a higher level of intrinsic sophistication and they certainly will be required to prove useful to compensate notably problems met by persons with reduced mobility. Nao, for example, can fetch an object and bring it back to his owner.
Robots like these provide answers to mobility issues and will become increasingly important as world population’s average age rises; in the long run, development of robots in this category could be favored by social security systems; if reimbursement of the cost to purchase/hire robot assistants is decided, this will open the way to mass production.
However, the scope of functions offered by Nao and his equivalents remains limited. Contrary to the wildest assertions, it is difficult not to say impossible to assess when service robots will be able to sit down on our family sofas. It is only through hindsight that we can really understand their nature: as yet not finalized research objects. The robot vacuum cleaner now on sale was in development for 20 years.
Google recently exuberantly announced the production launch of 100 driverless cars (with no steering wheels or pedals). We can bear in mind that the ‘majors’ in automobile manufacturing have been working on the driverless concept since the early 1980s and that the DARPA “Grand Challenge” is now 10 years old. This is a competition organized by the American Defence sector R&D Agency, for the purpose of obtaining driverless vehicles. Robotics is a costly and time-consuming field. To attain their objective, i.e., real reliability, there must necessarily be a long-range research approach. So, let us be realistic: personal service robots will not leave laboratories and drawing boards in the next 10-15 years.
If specialized personal service robots (vacuum cleaner, window cleaners, etc.) will continue to spread, the more generic service robots (personal companions, or assistants) have yet to overcome their basic shortcomings. From a motor point of view, the upright standing position has not yet been fully acquired. Mobile roller-mounted robots are still the majority. As far as perception is concerned, no domestic robot so far can recognize a generic object, such as a chair, while a two-year old baby can (chairs come in many shapes and sizes). Where handling properties are concerned, the manipulation skills are far from perfect. Removing an object from a table without breaking it, opening a door or drawer, folding linen or handling non-solid objects … are among the really hard things to (re)create in a robot form. RI-MAN, a robotic nurse created in 2006 by Riken, can lift and move a patient… but that is about the limit.
And if the motor functions are far from perfect, what about emotional aspects? There are algorithms that can mime laughter or empathy, but robots too can make mistakes. They can detect a human smile, but do not know how to interpret this. They are totally incapable, for example, of perceiving pain. In other words, robots do not yet have the capacity to imitate, or to adapt to (emotional) situations. As Pierre-Yves Oudeyer, senior research scientist at INRIA, framed it during a lecture he gave in November 2010: “For a robot, a room full of children is a far more hostile environment than any deep-lying ocean floor. Children will always try to invent new forms of interaction, new games … and the robot simply cannot respond.”
All told, decision taking by robots today comes down to applying pre-recorded responses to a symbolic changing world. Thus a robot can beat a world champion chess player, or book a train seat by phone. But for what we could call the domesticus socialus robot, the challenges represented by the real world remain almost complexly unsolved.