#	Pagina
attuale pagina	/open-h2020/projects/213161/results.html

Report

Teaser, summary, work performed and final results

Periodic Reporting for period 1 - MEMMO (Memory of Motion)

Teaser

Summary

\"What if we could generate complex movements for a robot with any combination of arms and legs interacting with a dynamic environment in real-time? MEMMO has the ambition to create such a motion-generation technology that will revolutionize the motion capabilities of robots and unlock range of industrial and service applications. Based on optimal-control theory, we develop a unified yet tractable approach to motion generation for complex robots. The approach relies on 3 innovative components.
1) a massive amount of pre-computed optimal motions are generated offline and compressed into a \"\"memory of motion\"\".
2) these trajectories are recovered during execution and adapted to new situations with real-time model predictive control, allowing generalization to dynamically changing environments.
3) available sensor modalities (vision, inertial, haptic) are exploited for feedback control which goes beyond the basic robot state with a focus on robust and adaptive behavior.

To demonstrate this approach, MEMMO is organized around 3 relevant industrial applications, where MEMMO technologies have a huge innovation potential. For each application, we will demonstrate the proposed technology in industrial or medical environments, following specifications designed by the end-users partners of the project.
1) A high-performance humanoid robot will perform advanced locomotion and industrial tooling tasks in a 1:1 scale demonstrator of an aircraft assembly.
2) An advanced exoskeleton paired with a paraplegic patient will demonstrate dynamic walking on flat floor, slopes and stairs, in a rehabilitation center under medical surveillance.
3) A challenging inspection task in a real construction site will be performed with a quadruped robot. While challenging, these demonstrators are feasible, as assessed by preliminary results obtained by MEMMO partners.

Memmo is a collaborative project that seeks a breakthrough in the use of complex robots in industrial and medical scenarios. Based on the definition of a new concept, the \"\"memory of motion\"\", it will lead to a new technology for robot control, that has potential to impact several market domains. The variety of impacts is emphasized by demonstrating the technology in 3 relevant environments, defined by our partner stakeholders: PAL ROBOTICS in Barcelona is targeting the market of mobile robots for end-users like AIRBUS; WANDERCRAFT in Paris has designed one of the most advanced exoskeleton for paraplegics, targeting, for the first time, rehabilitation centers like those handle by the Center for Physical Medicine and Rehabilitation of APAJH, based in Pionsat, France; and COSTAIN, a civil-engineering stakeholder, is seeking new solutions for inspection in its constructions sites. The objectives of the project require a collaborative joining expertise in motion planning (brought by LAAS-CNRS, Toulouse, France), robot learning (brought by IDIAP, Switzerland), computer vision (brought by University of Oxford, UK), force control (brought by Max-Planck Institute, Tubingen, Germany and University of Trento, Italy), optimal control (brought by University of Edinbourgh, UK), and capabilities to set up realistic pilot experiments in robotics.
\"

Work performed

The 1st year has been devoted to setting up the structure and the basis of the project. While the main collaborations inside the consortitum were bootstraped, we built a 1st prototype of a complete memory of motion for simple robots (inverted pendulum, quadcopters). This prototype, formulated as the IREPA algorithm, uses off-the-shelf components (the ACADO trajectory optimizer, a probabilistic roadmap for storing the optimal trajectories, a neural network to approximate and interpolate them) and demonstrated a quick and safe convergence up to dimension 15. The goal of the project is now clear: to scale this algorithm prototype up to dimension 100, the typical dimension of humanoid and other legged robots.

The consortium also released early versions of the final components of the memory of motion. In the 1st workpackage (dedicated to the generation of the basis of movements using motion planning), we released a locomotion planner and run it on a cluster to produce to locomotion database, totaling aroung 60Gb of motion data. In the 2nd workpackage (dedicated to the encoding of the motion database into a memory of motion using machine learning), we have explored several possible formulation to learn this mass of data, and decided which strategy we will implement during the 2nd year. In Workpackage 3 (dedicated to optimal control), we have implemented a optimal control solver for legged robot, and scale it expected performance to 10 ms of computation for 1 second of previewed motion of the humanoid robot: these performances go far beyond the state of the art and beat our optimistic expectations. In workpackage 4 (dedicated to sensor-based feedback), we have proposed a model to capture the sensori-motor behavior of a robot in contact based on measured data, that should later be exploited with the optimal control solver. In workpackage 5 (dedicated to embodiment on the robotic platforms), we have design a new version of our exoskeleton, implemented an excellent torque controller on the TALOS robot and prepared dense sensing algorithms that have been tried on the quadruped robots. Finally, the other workpackages have set up the experimental platforms and the benchmark criteria and prepare the scenarios and the demonstrators corresponding to the 3 industrial case studies: humanoid coworker in the factory of future, exoskeleton for rehabilitation and quadruped for inspection.

Final results

\"In the begining of the 2nd year, the consortium focuses on releasing a 1st complete memory of motion encoding the motion database and an efficient implementation of the optimal control solver. Based on these 2 technical achievements, we should demonstrate a 1st complete implementation of real-time optimal control on the real robots of the project by the end of the year.

The main idea is that we have a basic prototypes of the \"\"memory of motion\"\" concept, running for simplistic systems ; and the standalone implementations of all the part of the scale-up version of the memory, that we want to implement on our robots: we have a full-scale motion databased produced by CNRS with the help of Univ. Oxford, Univ Edinburg and IDIAP ; we have an implementation of the machine learning algorithms that we want to use for encoding the memory, produced by IDIAP in collaboration with Max-Planck Institute en Univ. Edinburgh ; we also have an optimal control solver with sufficient computation efficiency (developed by CNRS with the help of Max-Planck Institute and Univ. Edinburgh) and the sensori-motor feedback models to use it on the robot (developped by Max-Planck Institute with the help of IDIAP) ; finally we have the 3 robots ready and a definition of the scenarios exemplifying our practical case-studies: humanoid robot in an aerospace factory of the future, exoskeleton for rehabilitation in a medical center, and quadruped in a civil-engineering environment for inspection.

The work of the 2nd year is to put all these scale-up components in a common architecture, and to demonstrate the validity of the concepts in laboratory. Based on such a success, we would then bring our robots in the real industrial environments in the 2nd part of the project. See you in December 2019 for the 1st tests of the memory of motion on our 3 robots.
\"