February 18, 2010      

Aristotle might have argued against the concept of a “universal good,” but for roboticists “autonomy” would be a solid contender. The specific reasons why autonomy in robotic systems is regarded so highly vary according to the application. However, increased functionality is the reason typically given, often measured in terms of increased productivity in industrial settings and enhanced capabilities for military applications. Autonomy in robotic systems can also decrease the overall costs for certain robotics application solutions, although the relationship between autonomy and cost reduction is often overlooked.

Autonomy Is the Opportunity
That increasing autonomy is an overarching trend within all segments of the robotics and intelligent systems industry is beyond dispute. Again, the conventional wisdom among those who employ robotic systems for serious work holds that greater autonomy in robotic systems reduces application costs and increases system capabilities. In addition, the same technologies that support autonomous operation in robotic systems can be incorporated into products and services not deemed “robotic” by the general public, such as self-parking cars.

For all members of the robotics value chain, from researchers to technology providers, and on to end users, autonomy is the opportunity—the opportunity to secure research grants, drive revenue, or do a job more effectively. But a hard-nosed business analysis limited to the strategic and business advantages that increasing autonomy conveys to robotics systems, and the entities that employ them, provides only a partial understanding of the scope of the opportunities.

Autonomy in robotic systems will increase because robotic systems with growing levels of autonomy will be developed. It is a classic case of circular reasoning, but true nonetheless. In short, there is a generally held belief that over time autonomy will become pervasive in both civilian and military applications, robotic and otherwise. As technology improves, and autonomous systems build on their success to become commonplace, the number and scope of autonomous applications will multiply exponentially. As such, autonomy is a transformational capability, and therein lies the opportunity.

This article provides an overview of the ways that the increasing levels of autonomy is changing the robotics landscape across all industries, and how that may impact current robotics and intelligent systems technology providers, new market entrants, and the investment community. It includes a breakdown of the various classes of autonomous capabilities, and describes the role of autonomy in various vertical market segments—consumer robotics, military robotics, and industrial robotics—along with a review of enabling technology that is central to greater autonomy in all classes of robotics products.

Autonomy and Robotics
Autonomy can have different meanings for different people, even if the discussion is limited to the views of roboticists. For the robotics community, autonomy is not equated with independence, or degrees and types of self-determination. This is a completely different situation for the population at large. In the latter case, the term “autonomy” connotes freedom, sovereignty, self-determination, and self-organization, subjects of serious study for Aristotle (again), Kant, Spinoza, and other philosophers, while for biologists, psychologists, and social scientists, autonomy is suggestive of life, cognition, and self-awareness.

For the robotics community, as well as for the purpose of this article, autonomy denotes the capacity of a robotic system to move or perform tasks responding to both environmental and internal stimuli without external control. “Control” in the preceding sentence, however, does not include a power source, or the programming and engineering action of an external designer. These are a given, for at this time robots are not autonomous in the larger sense. Robotic systems require humans to supply them with goals, instructions, power, and other types of support.

Types of Autonomy
Even within the narrowed definition of autonomy as it applies to robotics and intelligent systems, there is disagreement as to what constitutes autonomy, along with the various forms that autonomy can take. However, there is considerable agreement on how to measure or scale autonomy; namely, the amount of operational time between instances of direct human input. Common classes of autonomy applied to robotic systems, include, but are not limited to:

  • Task and Mission Autonomy. Today, task autonomy refers to the ability of systems to complete a discrete piece of work, often involving the control and operation of end-effectors to achieve some limited objective, without external human control. In these instances, task autonomy and manipulation autonomy are nearly synonymous. In the future, as robots evolve to carry out more complex tasks, particularly where human supervision is impractical or impossible, it will become necessary for systems to offer some level of autonomous goal development, activity planning, and decision making. The long-term trend, then, is the evolution from simple task autonomy to the more expansive mission autonomy.
  • Power Autonomy. Power autonomy is often incorrectly referred to as the amount of time a robotic system can operate before its power source must be replenished or replaced. A more appropriate definition would include the ability of robotic systems to monitor and manage their power systems, and automatically seek and secure new energy sources when necessary.
  • Failure Autonomy. Failure autonomy describes the ability of systems to monitor their own hardware and software, detect problem conditions, and repair, recover, or failover gracefully.
  • Guidance and Navigation Autonomy. Guidance and navigation autonomy, the capacity of systems to safely move without external control, is the most common form of autonomy exhibited by robots at this time. This form of autonomy has a long history, with years of research in areas fundamental to the development of guidance and navigation autonomy such as path planning, mapping, localization, and collision avoidance.

Degrees of Guidance and Navigation Autonomy
A number of factors determine the levels of guidance and navigation autonomy exhibited by specific systems. The criticality of the system is a key determinant, along with the system’s potential to do harm (i.e., surgical robots versus robotic smart toys). Also, in many cases human input can optimize the efficacy and efficiency of semiautonomous systems. Guidance and navigation autonomy runs a continuum from teleoperation to fully autonomous behavior (see Figure 1) as described below:

  • Teleoperation. Dedicated, continuous remote operation without exception handling. A human operator makes all decisions.
  • Teleoperation with Exception Handling. Continuous remote operation with exception notification when problems occur (“wheels spinning, unable to proceed”).
  • “Directed” Autonomy. Systems directed to “go there” and “go there next” with minimal, noncontinuous, direct control. Operator guidance is provided when problems arise.
  • Autonomy with Oversight. Systems “go there” with no direct, continuous control by human operators, making path-following decisions by themselves.
  • Autonomous Operation. Complete autonomy, problem resolution, and correction capability (“go there, perform this task, and return”).

Product Classes
At one time, the three major categories of robotic systems could be largely distinguished based on the operating environment and general levels of autonomy they exhibited as a class. For example, consumer products such as robotic vacuum cleaners were designed for the unstructured environments found in homes, with many products exhibiting a high degree of power and navigation autonomy. Industrial robots—immobile, single-task systems under strict and simple programming control as they performed their tasks in a manufacturing or factory floor automation role—exhibited little autonomy and operated in highly structured work cells. Service robots, which include numerous subclasses ranging from military unmanned ground vehicles (UGVs) to hospital automation bots, fall between industrial systems and home systems on the autonomy/operating environment continuum.

As technologies and applications have advanced, all classes of robotics and intelligent systems technologies have increased the levels of autonomy they support (see Figure 2). Certain subsectors, however, better illustrate the trend, but more importantly, offer the greater technology, product, and investment opportunity.

Consumer Robots
Consumer robots, those robotics products purchased by individuals for use in the home, often display autonomous behavior, dependent, of course, on the particular class of product. The consumer robotics market includes the following classes of product:

  • Robotic smart toys
  • Educational robotics/hobby kits
  • Hobbyist robotics
  • Home-care/lawn-care robots
  • Home healthcare systems
  • Personal robots

Of these, smart toys, home-care robots, and home healthcare robotic systems can benefit the most from the inclusion of autonomous behavior.

[Editor’s Note: Personal robots, often called “companion robots,” “personal service robots,” or “network robots,” are technically advanced robots that interact directly with people and are designed to assist in the home, or to act as a companion, often to the elderly. Few personal robots have been released to the marketplace and these are almost exclusively prototypes.]

Smart Toys
Robotic smart toys are variations of classical children’s toys that have added sensing, intelligence, and mobility. Examples of robotic smart toys include Innvo Labs’ Pleo, Innovation First’s HEXBUG Nano, and WowWee’s Robosapien series.

Many robotic smart toys are controlled remotely (programmability is limited), but others exhibit some manner of autonomy. In fact, the addition of autonomous behavior is one method for overcoming the “one day of play” phenomenon where robotic toys are abandoned shortly after they are purchased. That is, autonomy increases the play value of robotic smart toys.

Autonomous behavior in robotic smart toys can take many forms, including:

  • Development. Products are “birthed” and then develop an autonomous social presence over time, including random behavior, following social interaction with the toy’s owner.
  • Learning. Smart toys learn over time, adjusting their “autonomous” behavior based on external stimuli. Both development and learning in smart toys are a limited type of task autonomy.
  • Mobility. Products exhibit intelligent mobility, particularly autonomous movement (navigation mobility).

[Editor’s Note: A fuller discussion of the various methods that can be taken to overcome the “one day of play” syndrome with robotic smart toys is provided in the article, “Selection Criteria for Consumer Robotics Products” in the January 2009 issue of Robotics Business Review.]

Home-care/Lawn-care Robots
Home-care/lawn-care robots are low-cost (mostly), single-function robots used in and around the home to perform household chores. Autonomy contributes to the two leading reasons why consumers would consider home-care/lawn-care robotic products: They save time and are convenient (see Figure 3). Robotic vacuum cleaners are an example. The first versions of iRobot’s Roomba, for instance, provided coverage through a combination of random movement and wall following. The procedure was basically autonomous, offering a fast and convenient way to vacuum a floor.

Following generations of the Roomba added the ability to automatically return to charging stations (power autonomy), as well as support for vacuuming at specific times and extending vacuuming based on sensing dirt quantity (task autonomy), again offering more convenience. Roomba competitors, such as Neato Robotics’ XV-11 and Samsung’s Furot II, also boast these features, but have gone further, adding navigational and mapping capabilities so that complete coverage is assured and areas are not revacuumed unless they require it.

Robotic lawn mowers such as Friendly Robotics’ Robomow, BelRobotics’ BigMow, Husqvarna’s Automower, and Kyodo America’s (KA) LawnBott, have followed a similar pattern of leapfrogging features, largely related to increasing navigation, power, and task autonomy. (See “A Look at the Robotic Lawn-Mower Marketplace,” page 1.) Purposeful navigation is common among robotic lawn mowers, and the automatic return to charging stations is standard functionality. Rain sensors that signal mowers to return to charging stations during periods of rain or when sprinklers activate are becoming standard. Other systems can also detect wet grass or grass height, forcing systems to retire if mowing conditions are not optimal. Like robotic vacuum cleaners, robotic lawn mowers can be programmed to mow at specific times. Some high-end models go further, offering a self-programmable mode where systems can calculate their mowing times based on lawn size and the growth rate of the grass.

Home Healthcare Systems
Many different products fall under the aegis of home healthcare systems. Those that are “robotic” are a subset of the more general assistive technology. Those devices that support the daily living functions of individuals through the use of robotic technology—mechatronic force or action based on information captured through sensor systems—define robotic assistive technology. Robotic assistive technology can further be subclassified into:

  • Robotic Mobility Aids. Mobility aids provide locomotion and navigational assistance to individuals with disabilities. Examples of common mobility aids include wheelchairs, walking aids, scooters, and power chairs. By adding robotics technology that increases autonomy, such as sensors, navigation technology, and intelligent power systems to traditional mobility aid platforms, products become usable by individuals who find it difficult or impossible to operate the traditional products.
  • Robotic Manipulation Aids. Robotic manipulation aids typically incorporate some type of robotic arm that is under the direct control of the end user. The devices allow individuals to more easily perform daily living tasks such as reaching, gripping, dressing, cooking, eating, and so on. The arms are usually attached to a wheelchair or a specially designed desktop. Research into increasing task autonomy (i.e., “secure cup and bring to lips”) robotic manipulation aids is ongoing.
  • Robotic Social Assistance Aids. Social assistance robots are a new type of robotic system specially designed to engage robots and humans in nonphysical social interaction. They are intended to be therapeutic aids, to increase the enthusiasm and compliance of patients during therapy, and as cognitive aids, to act as a reminder or assist with decision making and reasoning. Increasing levels of task autonomy make such products more engaging and functional.

Military/Security Robotics
Most developed countries understand the value of applying robotics technology to military requirements and have spent considerable time and money delivering operational systems that are some of the most advanced in the world. Unmanned aerial vehicles (UAVs), developed over a period of 40 years, have proven their worth, and as a result, countries are quickly adding unmanned aerial systems (UAS) to their inventories. The United States leads in this regard with the Department of Defense (DoD) increasing its number of unmanned aircraft from 167 to more than 6,000 from 2002 to 2008. At this time there are approximately 5,000 ground robots employed by the U.S. military (more than 13 systems), up from less than 100 in 2001. Improvised explosive device (IED) detection, reconnaissance, and explosive ordnance disposal (EOD) are typical applications. Robotic systems for surface, subsea, and littoral operations are also in development.

Enabling Capability
At this time, robotic systems for use in the military are almost exclusively teleoperated. Even with this limitation, the systems have proven very effective. As a consequence, new and expanded capabilities have been called for by the world’s militaries.

Autonomy is considered a critical enabling capability for the next generation of military robotic systems. Much research and development is currently being undertaken to increase system autonomy. These efforts target all classes of unmanned systems, across all levels of autonomous behavior, for all pertinent technology and techniques. That is, increasing autonomy is a universal theme across all advanced militaries and among all suppliers of unmanned technology that serve them.

More Than Productivity and Casualty Reduction
Greater autonomy benefits military robotic systems in many of the same ways as their civilian counterparts. Yet in many respects, the payoffs for military systems are unique. In general, the benefits of autonomy for robotic systems for military duty fall in the areas of:

  • Operational Effectiveness. Similar to civilian systems, autonomy increases the operational effectiveness of military robotic systems on the battlefield (the functional equivalent of increasing productivity in industry).
  • Cost Reduction. To date, cost reduction has not been a driving force in military robotics acquisition programs. First, the individual robots are expensive. Second, the employment of robots has not resulted in a reduction in the number of military personnel. In fact, they increase the number of personnel as the robots themselves require operators and maintenance teams. But as the robots become more autonomous and functional, the maintenance load will drop and the systems can be applied in an increasing number of areas. It can be seen, therefore, that advances in autonomy, functionality, and robustness will not only act to increase mission capabilities, but will also reduce overall system costs.
  • Causality Reduction. Unlike civilian systems, military robotics can add casualty reduction to increased effectiveness and cost reduction as the primary reasons for embracing robotics technology. Robots can be used in place of humans for many of the dangerous and life-threatening tasks that soldiers perform on a daily basis.
  • Human Limitations. Military officials in the United States and elsewhere are expecting even more from future unmanned systems, requesting that research focus on what robots can do better than humans, as opposed to what humans already do well. In addition, with the teleoperated military systems of today, decision making and control are limited to human response times, often measured in minutes versus the subsecond responses of computer systems. Future combat scenarios might have unmanned systems in conflict with other unmanned systems. In such cases, unmanned systems with humans in the loop for tactical awareness and control might be at a disadvantage against more fully autonomous platforms.
  • Communication Limitations. Teleoperated robots are typically controlled using tethering systems or by using either direct radio frequency (RF) or satellite communication links. Tethered systems are bound by the length of feed line running from robot to control stations. RF- or satellite-controlled systems are much less restricted in terms of distance between controller and robotics platform. Still, link allocation and bandwidth issues, along with the possibility of jamming, are drawbacks. Increasing levels of autonomy, perhaps coordinated with low-bandwidth, asynchronous, event-driven control, can overcome some of the communication difficulties while the systems evolve over time to become fully autonomous.
  • Collaboration. One area of autonomous behavior that is of particular importance to military planners is multisystem collaboration, among multiple unmanned systems, as well as groups of manned and unmanned platforms working together. For example, teams of UAS and UGVs can work cooperatively to share situational awareness and distribute sensing. The speed of decision making required for systems to collaborate effectively during large-scale operations exceeds that of human operators. The only effective means of controlling and coordinating multiple robotic systems is to have many behaviors execute autonomously without external control.

Industrial Automation
Both the major industrial robotics companies, which are at risk through their overdependence on the automotive, electronics, and semiconductor industries, and smaller firms hoping to enlarge their businesses, are working to expand the use of industrial robots beyond their traditional industries and functions. Many of these initiatives focus on providing automation solutions to new vertical market segments, such as the food industry. Other efforts involve automating nontraditional roles such as packaging and warehouse management.

The expanded use of industrial robots beyond their traditional industries and roles requires new, innovative forms of enabling technology, much directed at increasing levels of system autonomy. Three areas of innovation in the industrial automation sector at this time include:

  • Complex Assembly. Efforts are under way to automate whole new classes of complex assembly applications that were previously limited to humans. These systems are designed to perform a wide range of manufacturing tasks that demand a level of dexterity, manipulation, and fine control lacking in traditional industrial robotic systems. Such systems must exhibit a high degree of task autonomy (manipulation autonomy), combining input from force, touch, position, and tactile sensors, along with machine vision, with intelligent software designed to solve problems and learn from experience to solve high-order tasks.
  • Human-Robot Interaction. A new generation of human-scale industrial robots is under development, engineered to work in close association with humans safely and efficiently in unstructured or semistructured environments. Guidance and navigation autonomy, along with advanced forms of task autonomy, are required to realize this new class of robotics workforce helpmates.
  • Automated Warehouse Management. Distribution centers and warehouses continue to increase their levels of automation to reduce costs and improve productivity (the usual), but also to increase their flexibility. A progressive increase in the levels of navigation and guidance autonomy, along with power autonomy, will play a key role as autonomous mobile vehicles begin to replace automated guide vehicles and manual methods for movement, storage, and retrieval of goods and material. For certain applications, guide-based systems operating in rigid, highly structured environments, will give way to perception-based systems able to navigate in dynamic, unconstrained environments.

Technology and Research Opportunities
An exhaustive list of every technology and area of research required to support increasing levels of autonomy, across all classes of robotics systems and applications, is beyond the scope of this article (and probably impossible). However, the greatest areas of opportunity are linked to their respective impact on some aspect of autonomy and the degree with which functional gaps act as a bottleneck. Common areas of opportunity applicable to all domains include:

  • Task Autonomy. Perform complex tasks without human control and interaction using:
    • Fine Control and Dexterous Manipulation. Ability to coordinate tactile, vision, and proprioceptive sensing to manipulate objects with human-like dexterity
    • Sensing. Increasingly rugged sensors and sensing systems (including multirobot sensing), reduced in size and power consumption, which provide high resolution and signal quality.
    • Perception and Situation Awareness. Ability to base control on perception and situation assessment, based on input and integration from a variety of sensors including vision, touch, inertial, and odometric.
    • Object Recognition. Capacity to visually recognize and categorize objects using model comparison, pattern recognition, edge detection, and other methods.
    • Adaptation. Ability to react and adapt to unexpected situations.
    • Machine Learning. The capacity to learn while engaged in the problem-solving process, as well as to learn from previous experience.
  • Power Autonomy. Overcome power source limitations using:
    • Advanced Power Supplies. Availability of lightweight, long-endurance, high-density power supplies including advanced battery systems and fuel cells.
    • Power Management. Software that monitors power consumption and adjusts functionality to optimize.
    • Energy Harvesting Technology. Capacity to generate power through the conversion of biomass and sunlight, or by using petroleum-based products, wall and wire electrical current, and other “found” sources.
    • Autonomous Power Charging. Support for returning to charging stations before power levels are depleted.
    • Standard Charging. Ability to locate and utilize standard power outlets as power sources.
  • Failure Autonomy. Automatically recover from an error state through:
    • Self-Monitoring and Diagnosis. Technology that can perform fault detection and recovery using rules-based and fuzzy logic.
    • Hardware and Software Redundancy. Support for systems failure by way of redundant hardware and software systems, as well as software partitioning.
  • Guidance and Navigation Autonomy. Navigate environments autonomously using:
    • Obstacle Detection and Avoidance. Ability to avoid both fixed and moving obstacles through tracking, reactive control, and other methods.
    • Navigation and Localization. Navigation in dynamic, unconstrained environments using simultaneous localization and mapping (SLAM), terrain navigation, global positioning system (GPS), and other techniques.
    • Intelligent Route Planning. Ability to determine the optimal route based on models and range of sensing modalities.
    • Advanced Communications. Robust, long-range communications, using RF, infrared, and ultraviolet, capable of withstanding environmental interference and jamming.

Conclusion
It is widely held that increasing levels of autonomy reduces the costs and increases the capabilities of robotics systems. Other benefits are also conferred based on the specific application. Even without the benefit of these truisms, it is widely understood that all classes of robotic systems—ranging from consumer bots to military systems—will become more autonomous over time. Eventually these systems will reach the point where the majority of applications will operate without the need or benefit of human input. This represents a truly transformational change. So much so that in certain quarters, it is “autonomy” and not “robotics” that is considered the real opportunity. In either case, this inevitability presents a massive opportunity for all members of the robotics value chain.

The Bottom Line

  • The definition of autonomy in the context of robotics systems differs from traditional characterizations in that it assumes goals, instructions, power, and other types of support are provided by an external designer.
  • Degrees of autonomy are measured as the amount of operational time between instances of direct human input.
  • Autonomy is a transformational capability, and as such, increasing autonomy is an all-encompassing theme acting on every segment of the robotics and intelligent systems industry.
  • As technologies and applications have advanced, all classes of robotics and intelligent systems technologies have increased the levels of autonomy they support.
  • One long-term trend for robotic systems is the evolution from simple task autonomy to the more expansive mission autonomy.
  • Guidance and navigation autonomy takes many forms, ranging from teleoperation to fully autonomous operation, based on the level of human support.
  • Opportunities in autonomy are almost too numerous to note, a reflection of the wide range of applications that benefit from the capability.

Resources