Robotics is a typical transversal domain that brings together several very different scientific and engineering areas going from mechanical design of automatically controlled, multipurpose, manipulator to multi-modal emotional man machine interface. Most Digiteo partners may therefore contribute, directly or indirectly, to fundamental or applied research in robotics. Natural links have to be developed with activities outside Digiteo, in Sciences of System Engineering, Physics, Social and Human Sciences.
The main characteristic of a robotic system is to be able to autonomously adapt its behaviour to its environment. The robot needs to seamlessly perceive its environment, compute the most relevant behaviour according to its perception, its memory and its goals, and act. Digiteo partners conduct research along these axes :
Modelling and design for robots components, control and interactions :
- Design of master arms, slave robots for hostile environments, collaborative robots for production, health applications, and mobile manipulation.
- Modelling and identification of complex mechanical structures with the objective to characterise their rigid body motions as well as their mechanical flexibilities, and to enhance the friction’s modelling for back-drivable mechanisms or long-range carriers.
- Development of appropriate man-machine interfaces based on haptic devices and graphical interfaces
Perception is needed to localise a mobile robot, to build a map of its environment and to identify objects, obstacles and people within this environment. Digiteo actors are involved in :
- Localisation and mapping based on inertial sensors, odometry, satellites, computer vision and distance sensors. Various approaches are applied, from low-level data filtering to high-level semantic mapping
- Bio inspired methods for localisation and object recognition.
- The study of perception-action relationship and active perception (e.g. attentional mechanism) in which action is required for perception.
- Acoustic, prosodic and lexical markers detection (based on audio signal features extraction, speech recognition and machine learning such as SVM, Markov models) and multimodal perception (visual cues extraction, fusion of audio and visual cues) for emotion detection in Human-Robot interaction
- Haptic perception with and mainly without force/torque sensor dedicated to master the interaction between Human and Robot and Robot in its environment in terms of force interaction.
Path planning and exploration with different methods,
- 3D supervision of systems based on real time linking of robot and interactive simulation.
- Learning for the long-term adaptation of robot to their environment and adaptation of their behaviour to the user expectations.
- Defining self-driven goals conducive to autonomous robotics, for instance in the context of robotic swarms.
This issue requires efforts on the design of the physical architecture of the robot.
- Low level control loop of classical electrical actuators or new technology based actuators (shape memory alloy, piezo-electric transducers, magneto-rheological fluids...) using conventional approaches as force or impedance control but guarantying robustness properties with respect to modelling errors or unmodeled dynamics, or learning and evolutionary methods,
- Adaptive control architectures and models of the perception-action loop based on advanced automation methods and connectionist, bio inspired and Bayesian paradigms
- Development of vision based control methods to control either mobile robots or manipulators in order to servo robot motion on vision feedback,
- Stabilisation techniques robust to communication and transportation delays, constant or time-varying, distributed or not over some time-intervals.
- Safe and reliable object manipulation for service robotics.
Based on the technologies presented above, the main applications developed by Digiteo partners are : flying robots, mobile robots, robotic swarms, automatic car control, tele-robotics, adaptive robotics, collaborative robotics, assistance to disabled people, cognitive and physical stimulation to elderly and disabled people, haptic interfaces, humanoid robot and miniinvasive robotic devices.
One important issue for the large-scale development of robotics is safety. Humans and robots will have to work together closer and closer. It is necessary for robotic industry to ensure that robot will not become a threat for human beings. Ability to prove that software is doing what it is designed for is critical. The software engineering expertise developed by Digiteo partners can be applied with great benefit to robotics.
Current Digiteo initiative
Digiteo partners therefore offer a large expertise in many of the area involved in robotics and related problems, from the fundamental mechanics to the emotional modelling and from the generic design to the final integration of industrial prototypes.
A first Digiteo initiative called the ROBOTEO project aims at building a federative robotic platform for Digiteo research laboratories. 3 themes were selected in different application and scientific domains but presenting complementarities : Robotic assistant, manipulation and autonomous vehicle. The Roboteo platform will make up with an electrical vehicle, a force feedback master arm, a slave robot, a collaborative robot (or Cobot), an assistant service robot and open generic robot control software. This platform will be used as demonstration devices for scientific projects or as training supports.
The Roboteo platform is currently under implementation ; materials were selected and call for tender are under progress for master arm and a cobot. An existing slave arm and a haptic interface are also currently dedicated to scientific programs.
Two PhD theses started in 2010 between Digiteo partners focusing on control of robot and remote handling systems. Two other Research programs are planned to start in 2011 focused on cobot control and autonomous manipulation.