Ten Top Roboticists at Collège de France, Paris
As we announced in a former post, the Collège de France hosted in Paris a two days seminar with 10 among the world’s most prominent roboticists. For those who couldn’t make it, the videos are available online. We provide below the links as well as a short summary of each talk.
Pr. Jean-Paul Laumond: Welcome Address
In this introduction, Pr. Laumond reminds the history of 50 years of robotics and discusses the question whether robotics is a science or a technology. He also describes the current state of the art and lists major expectations of the field:
- Mars and more generally space exploration
- Medical applications: surgery, nano-robots/robotic particles for cancer healing
- Operating in dangerous situations as targeted by the Darpa humanoid challenge
- Exoskeletons especially for the disabled
- Ubiquitous networked robots to assist humans or for surveillance
Pr. Paolo Dario: Science and Technology of Robot Companions
Pr. Paolo Dario’s talk is about milestones of bio-robotics research towards robot companions. First, Pr. Dario provides a definition of bio-robotics. Then, he summarizes his own research history with a very impressive list of projects and cool videos. It covers tactile sensors, robotic hand and arm. He also conducted research on applications of mobile robots such as the Dustbot project. The latest projects of the impressive list described by Pr. Dario are related to robots for surgery and medical applications such as the Vector project which objective is to develop robotic capsules for endoscopy.
Pr. Mitsuo Kawato: Manipulative Neuro-Science. Control of brain-robot-world loop
The talk of Pr. Kawato is structured in 4 parts:
- Part 1 is about computational neuroscience and aims at understanding the brain by creating an artificial one based on biological facts. To make an intelligent brain, a body is required (embodied intelligence).
- Part 2 is about brain-machine interfaces. The major application targeted is to control an exoskeleton mainly for rehabilitation purpose.
- Part 3 addresses manipulation of neural code. It discusses how to induce activity of visual areas of the brain.
- Part 4 presents the network BMI project. The goal of to control a house equipped with many actuators and sensors using non-invasive Brain-Machine-Interface (BMI).
Sir Michael Brady: From Vision to Robotics, to Cancer, and back again
This talk is on vision systems and image processing for robotics but also for other applications and more specifically health and medical. Sir Brady explains that his personal driver is to positively impact people’s everyday life. This is why after several years working on UGV capable of indoor navigation with real time obstacle avoidance, he start exploring applications of his research beyond robotics: Cancer image analysis, Tags to track Criminals (instead of keeping them in prison), Dynamic boat positioning to help keep them at fix distance near an offshore oil rig in all weather conditions… All these applications have in common that they can be dangerous for humans and that the signal/image processing needs to work with 99.9% accuracy.
Pr. Russel H. Taylor: Medical Robotics and Computer-Integrated Interventional Medicine
Pr. Taylor starts his presentation by explaining that he considers invasive medicine from an information flow perspective because of his background as a former IBMer. He presented some of the challenges he faced in his work when using robots for micro-medical eye surgery:
• Small tools 1mm section
• Very little force feedback, which is less than what humans can feel
To address these problems, he used information technology to improve the process for every single patient, while at the same time he gathers statistical data that can used for all patients.
This result into augmented reality displays that combine multiple sources of information including vision, ultrasound sensors as well as optical fiber force sensors.
Pr. Hirochika Inoue: A Lifelong Challenge to Create Intelligent Robotics
This first talk of the second day of the seminar is a retrospective of research conducted by Pr. Inoue since the 1960’s. It is illustrated with many interesting videos. The presentation is organized in 5 parts:
- The first challenge addressed by Pr. Inoue is to build a manipulator with two fingers. This hand was equipped with a force sensor and could turn a crank, insert objects in holes, and stack objects. This is impressive especially knowing the limited sensors and computing resources of the time when the project was performed (1969).
- The second part of the talk is about the study of intelligent connection of perception to action. This led to building in1973 at the MIT a robot that performs fine assemblies. Studied hand-eye systems used Stereo vision and real-time 3D optical flow to guide the robot motion. It has been applied to wheeled and legged robots.
- Part 3 is about humanoid robots. Pr. Inoue has been involved in the development of different humanoids that eventually led to the HRP series.
- Research on humanoid robots including HRPs has been conducted as part of different Japan national projects. Different applications includes tele-operated humanoid driving a caterpillar, and robots for space exploration with the moon as a first target. An original idea was presented here. Instead of building rugged robots to endure extreme temperatures and lack of atmosphere (for heat transfer), the plan was to make humanoid robots wear space suites, similar to ones worn by astronauts.
- The last grand challenge is about robotic fleet that intervene on severe accidents or disasters. The goal is build humanoids smaller than humans (80% the size of a human) while stronger (150% to a human). These robots can be used as avatars for human manipulators that drive them remotely.
The presentation ends with a list of challenging tasks that the robotic community has to address.
Pr. Matthew T. Mason: Robotics Manipulation Uncertainty
This talk begins with the explanation of Pr. Mason’s point of view in favor of minimalism in robotic manipulation. That is the pursuit of the least complex solution to make robots do what humans can do. Many problems relative to manipulation are illustrated by videos of humans and apes such as:
• Dealing with more than one object per time
• Pushing objects one with the other
• Combination of grasp and holding different objects with different fingers
• Use of more than one hand to manipute a single object
• Wearing clothes
Pr. Mason’s noted that in many existing works, robots must have a correct model of the world. This model is updated upon changes that are sensed with certainty. He proposes to some of these assumptions on knowledge, and illustrate this approach with multiple examples from work he conducted.
Pr. Rodney Brooks: From Robotics Research to Large Scale Robotics Applications
This talk was given by Pr. Brooks that left iRobot and co-founded Heartland Robotics. This latter was renamed Rethink Robotics and released recently the Baxter manipulator robots for manufacturing applications. Pr. Brooks discussed how to turn research ideas and solutions into robots that can be useful for real world applications. He draw an analogy with computers.
Historically they were locked in rooms. Now, people can touch them and have many of them. The same applies to robots. First robots in manufacturing were isolated because they can injure, or even kill people. What if people could touch robots? It’s time to start robotic businesses today because of:
• Consumer pull : rehabilitation, dull tasks/jobs
• Technology push : Moore’s law, and available solutions such as SLAM and machine learning
Roboticists should be careful about the consumers/users that have different attitude. Pr. Brooks reported that some users put Roomba’s virtual wall on top of the robot. It is because of similar reasons that the roomba control panel evolved to a single button in the latest version. Similarly, iRobot replaced the advanced HRI initially developed for the PackBot and other military robots by a game controller. This latter has half capabilities of the original interface, but it eases adoption, since the device was already known to users.
Pr. Yoshihiko Nakamura: Robotics Technology and Human Science
In this talk, Pr. Nakamura discusses humanoid robots. The fundamental quest of research in this domain is to understand and support humans. One of the main challenges is communication. This applies to both interaction/collaboration between humans and robots, as well as for teleoperation, were the amount of data to exchange can be reduced by using higher-level information.
To address this challenge Pr. Nakamura proposed solutions inspired by mirror neurons. These neurons fire both when someone performs an action, and when he watch someone else performing the same action. The behavior of these neurons can be mathematically modeled using a Hidden Markov Model (HMM). Pr. Nakamura presented multiple advanced simulations he has been carrying. They show realistic control of muscles to perform various actions.
Pr. Gerd Hirzinger: Advanced Robotics and Mechatronics – From Space to Terrestrial Assistance, Surgery and Mobility
In this final talk of the seminar, starts by a presentation of Pr. Hirzenger research on robotic arms. He contributed to the first tele-operated robot sent to space in the Columbia space shuttle. It was controlled from earth by means of high-level commands. Optimizations of motor design lead eventually to soft robotics, where safety of human near robots is ensured.
Design of robotic hand also was also significantly improved. Multiple iterations on the design of the DLR hand lead to the first european robotnaut hand. It is robust enough that it cannot be damaged even after being hit by a hammer. This hand is mounted on the Justin robot that can now catch balls thrown by a human. Justin can move both the whole body and his hands according to the ball’s trajectory.
The talk ends with current research directions. Some are related to improving the robotic hand or arm. Others are about flying robots that may have a hand to perform manipulation from air. Another notable project currently conducted aims at sending 4 robots to the moon by 2014 as part of the Google’s lunar X Prize. Three among the 4 robots of the fleet will be rovers which the first version is already available. The fourth robot will be either yet a rover or a humanoid robot.