Georges Giralt LAAS-CNRS, Toulouse, France
Robotics opens today, at the turn of the Century, a large perspective for seminal scientific and technical achievements well articulated to a highly challenging broad host of novel applications.
At the theoretical level, Robotics emerges as a scientific body of concepts methods and algorithmic tools, in fact the most challenging field in Machine Intelligence which effectively interplays with a current stream of developments that pave the way at the practical level to a very large domain of novel applications ranging from Outer Space to Assistive and Personal Robotics.
The corresponding research trends and application developments are supported by the explosive evolution of enabling technologies: Computing Power, Telecommunications, Networking, Sensing Devices, Mechatronics, New Materials, Micro and Nano-Technologies, ...
Industrial Robotics will largely remain a well established domain in constant progress with a large variety of market products whose development trend will rightly blend the research issues within the framework of Shopfloor Automation.
The frontline aspects of Robotics Research appear to be set in the framework of what used to be called during the last decade "Non-Manufacturing Robotics" where the robot moves out of the well-engineered production-line environments to the vast spectrum of applications pertaining to Field, Service, Assistive, and Personal Robotics.
Salient features of what can be considered as the birth of true Robotics [rebirth of Robotics ?] stress the paramount role, in its various themes, of Machine Intelligence endowing the robot to act:
As a human surrogate in particular for intervention tasks in remote and/or hostile environments,
In close interaction with humans and operating in human environments in all applications encompassed by Human-Friendly Robotics,
In tight synergy with the user expanding into Human Augmentation.
In current and, even more, in future realizations endowed with Machine Intelligence, features such as Advanced Sensing and Perception, Task Reasoning and Planning, Operational and Decisional Autonomy, will more and more justify the name of Intelligent Robots.
These novel robots will be built in material structures, shapes, with sensors and end-effectors, the best suited for a given task and environment. The large variety of forms will certainly span from classical industry-oriented machinery through every kind of rover to human-friendly design concepts including anthropomorphic ones.
In the following, we attempt to categorize some of the novel application domains that can be considered as important and highly relevant of present trends and future perspectives.
Humankind frontier domains: encompasses applications of intervention robots for hostile and/or remote sites such as planet exploration, subsea operation, Earth difficult environments such as Antarctica, volcanoes, ...
Underwater Robotics which is already an economic reality mostly by means of ROVs (Remote Operated Vehicles) sees a raising interest for Autonomous Underwater Vehicles (AUV).
Robots for outer space able to operate in the highly non-cooperative environments of Mars, the Moon and other celestial bodies, will explore, carry on scientific experiments and build the outposts for humans to conquer the solar system.
Field-based applications such as mining, tunneling, dangerous waste and artifacts handling, forestry, agriculture, ...
Mining is a long-standing challenge for robotics in relation to harsh and dangerous work conditions.
Many places around the world are plagued with highly dangerous waste depots, or worse and most regrettably, scattered around in not well identified areas, anti-personnel mines that would kill, if nothing is done, hundreds of thousands of peaceful and hapless civilians.
This vast domain includes all sort of drones, cars, public-safety machines, professional cleaning, construction and civil work robots, surveillance, sewer inspection and maintenance, warehousing and inter-modal transportation, professional servicing and catering.
Remote task execution by users and expert monitoring and assistance will build on telecommunication, networking and Virtual/Augmented Reality interfaces for a host of activities such as plant maintenance, servicing, medical assistance, health care, ...
This is a fast developing category that offers both multiple highly difficult challenges for Robotics Research and whose economical impact appears to be of the utmost importance.
To its broader meaning the domain encompasses a variety of applications with significant differences in key research issues but all emphasizing the central problem of Human-Robot Interaction with modalities ranging from Human-Friendly Control Interfaces to Human-Machine coexistence and the synergistic aspects of Human Augmentation.
These application cases can be further classified taking into account in particular the interaction modalities.
Entertainment: toys, artificial pets, games, educational robots, ...
It is worth to underline that the two first sectors correspond to an actual beginning of market penetration.
The potential vast games area sees a quite important activity (Soccer, Tournaments) with an interesting impact on technical education.
Public oriented services: public places servicing (airports, museums,...), hospitals delivery, surveillance/safeguading, ...
First current results deserve to be mentioned:
some market opening for hospital delivery robots,
a number of relevant and successful research experiments with one or more robots as museum tour guide(s).
Assistive and Personal Robots
This challenging compound new field comprises cases such as Medicine, Rehabilitation, Household (cleaning, surveillance,...), Assistance to the impaired or elderly (companion robots, effective servicing, care, ...).
The last subject represents an utmost socially demanding and technically difficult research and development endeavor.
Both Assistive and Personal Robot application domains very well enlighten the concept of Intelligent Machine.
Human Augmentation
The field expands from Assistive and Personal Robots concepts, and somewhat merges with trends and issues raised in Computer Sciences (Neuro-Informatics) and Engineering (smart rooms,...). The robot breaks down into distributed functionalities to achieve Human Augmentation in aspects related to physical appearance, sensory capacity, information processing.
There is, of course, a strong continuity in research issues between Robotics "before" and the coming of "true Robotics".
Nevertheless, to consider applicative subjects such as an autonomous Mars Rover or a Personal Robot to assist an impaired person, leads necessarily to revisit trends and issues, to put at the proper place with the proper perspective the research topics and to bring some questions out of oblivion where the community had left them.
The following sections attempt to briefly scan some of the main and well-tended directions with the aim to comfort their interest in the light of Human Friendly Robotics chosen as the most novel framework. Furthermore it will be contended that there is a grand new challenge to be met for Robotics Research: to put at the top of its priority list Robot Dependability in Human Environments.
Indeed the specific needs introduced by the large variety of tasks and environments ranging from field applications to personal robots provide open opportunities for novel design concepts in all shapes and appearances building on new materials, telecommunications, networking, micro/nano technologies:
teleoperated and network connected machines,
multi-robot and multi-bodied systems where a sophisticated mother-machine will control work carried on by task-adapted simpler offspring robots,
legged and multiple-limbs task-adapted robots,
humanoids.
Some of the above design concepts lead to revisited control approaches. This is in particular the quite difficult case of locomotion and posture control for Humanoids.
For all, there is the imperious need to consider sensor-based control with multi-sensory inputs (force, contact, ranging, pattern recognition,...) to effectively relate motion and task execution to environment and task states.
i. Perception, Modelling, and Decisional Systems
The following topics are to be pursued in the right perspective to the forefront of Robotics:
Effective perceptual functionalities : vision (3D, color, texture), haptic,extended sensing (sound, odor, temperature, ...) and sensory-motor loops.
Extended modelling and data storage facilities: environment and tasks, general data bases.
Decisional autonomy: task reasoning and planning, multi-robot coordination, co-autonomy,control architectures.
ii. Friendly Human-Robot interfaces
Largely based on graphics to build virtual and augmented reality interfaces and using multimedia communication/control modalities (haptic, voice, text, ...) they will have to support key features such as:
User commands/intend understanding
Effective access to robot commands interpretation/actions
Machine constrained and imperative commands
iii. Learning
Learning will be undoubtedly a frontline research theme to develop the functionalities endowing public-oriented service, assistive and personal robots to efficiently operate in all kinds of human environments from public places to home.
Classical topical issues will call for extended efforts to achieve operational robustness and efficiency:
environment understanding, localization and motion control,
generalized learning by showing: tasks, object identification, ...
skill acquisition: object handling, cleaning modalities, ...
adaptativity to user communication facilities. This implies to consider the very difficult subject of the creation of user models.
The application cases encompassed by public-oriented service, assistive and personal robotics emphasize the human-machine interaction where the person may be either a non-professional user or a by-stander, or both.
This clearly points out to questions of Physical Safety but also to operating robustness in presence of wavering uses and downright misuses.
This shows clearly the utmost relevance and importance to integrate within design and software concepts, the key features corresponding not only to physical safety but in a broader way to the more general aspect of Robot dependability.
Although very little had been done so far on all the related issues in relation to the fields considered here, some preliminary work on autonomous robots began to deal with robot decisional architectures and exception cases handling. Currently one can foresee that exception cases will be the regular ones for Personal Robots, ...
The following briefly enumerates the main research issue for the theme of Robot Dependability:
Physical safety strongly builds on mechanical design and joint control for human interaction: soft materials, limited torque/force, speed, ...
Decisional and system dependability (algorithmics, software, sensing,...): reliable/failsafe sensing and control, diagnosis, exception cases handling
There is obviously a limit in the capacity to reach characteristics - certainly lower than 100 % - corresponding to the above contents in general terms of Robot dependability.
Clearly, we should have in mind the legal aspect but still focus our efforts on technical issues.
Once more, we should stress the extreme difficulty of several of the research themes involved in this many folded complex subject.
Several among the current research trends have not been specifically mentioned although they focus a reasonable amount of work by a fraction of the research community.
Subjects such as Swarm Robotics, Artificial Life, have momentum of its own but play a marginal role in the highly competiting and seminal technical arena we consider articulated with the many decisive application oriented directions of the frontline robotics.
Among those directions, it can be contended that there is an edge-cutting field within the larger domain of Human-Friendly Robotics: Public-oriented Service, Assistive and Personal Robots.
This field opens a double challenge in relation to:
societal and economical impact
in the interplay of two broad research themes: Machine Intelligence Based Systems and Robot Dependability.
The coming of this class of intelligent dependable robots is an actual grand challenge that brings into the real world some of the science fiction themes.
One can dream of Assimov's laws or may be of some specific implementation of part of them.
We will contend that research work during the next couple of decades should focus on efficient engineering implementation of Machine-Intelligence bounded rationality to devise Human-Centered Intelligent and Dependable Robots.