Final Report Summary - EXPLORERS (EXPLORERS Exploring epigenetic robotics: raising intelligence in machines)
The central target of EXPLORERS is to study how one can build robots capable of life-long and autonomous learning of novel sensorimotor and social skills that were not specified at design time. Central to the approach is the robot transposition and modeling of mechanisms that allow human infants to develop, with an emphasis on curiosity and imitation applied jointly to action and language learning.
This research program, framed in the field of Developmental Robotics, focuses on a selected set of developmental mechanisms known to be key in human development (intrinsic motivation, maturation, imitation, joint attention, properties of embodiment), with the primary goal of studying how they can be functionally formalized and transposed from humans to robots and explore how they can allow robots to be equipped with important aspects of developmental capabilities observed in humans. Besides this fundamental question, the combination of these techniques with classical robot engineering techniques, and their application to the new field of personal robotics, where robots need to adapt to non-expert users in everyday environments, is studied. A secondary goal of this research program is to use those computer and robotic models as tools for understanding particular aspects of human development.
Several key mechanisms allowing robot life-long and autonomous learning of novel sensorimotor and social skills were elaborated and experimented.
First, mechanisms of active curiosity-driven exploration, inspired by human intrinsic motivation systems, were devised to allow a robot to efficiently discover how to use its initially unknown body, and the interaction with external physical objects. This includes learning skills such as omnidirectional legged locomotion, arm inverse kinematics, or controlling external flexible objects such as a fishing rod.
Second, mechanisms inspired by human maturation were elaborated, allowing the robot to adaptively freeze and free novel degrees of freedom in both motor and sensori channels. This was shown to allow exploration and learning in large unbounded sensorimotor spaces, and to considerably speed up active learning techniques.
Third, it was explored for the first time how open-ended curiosity-driven learning of multiple sensorimotor tasks could be combined with imitation learning. These mechanisms allow robot to both set their own goals and try autonomous learning to solve them, and at the same time to benefit from human advice towards the interest of certain goals or to benefit from examples of strategies to achieve them.
Open-source 3D printed Poppt robotic platform for open science and education. The Poppy platform, developed within the ERC project, is the first complete open-source 3D printed humanoid platform in the world: http://www.poppy-project.org
The Poppy platform was initially designed with the goal of developing an experimental platform allowing the systematic scientific study of the role of morphology in sensorimotor skill acquisition (morphology can be changed easily and fastly). This has been a major innovation: this was before impossible because robot platforms were developed using classical machining techniques requiring a lot of time, energy and funding. Several research labs in Europe have already began to use the Poppy platform for their own project.
As all aspects of the platform were designed to be highly modular, modifiable, robust, easily replicable, cheap, and accessible to beginners, this allowed us to apply and initiate its transfer towards education in lycées, engineering schools and universities, but also FabLabs, science museums and art/science projects. It is now becoming a major pedagogical innovation, targeting education to the digital world and its interaction with the physical world through integration of the robotic platform with software web tools to support the community and with interdisciplinary pedagogical content co-designed with users. No direct competitors exist so far.
Poppy has been presented in highly visible and prestigious wide audience venues (François Hollande at Elysée, French Sénat, Le Web conference, Tedx Cannes), in numerous high quality media/press articles (e.g. Le Monde, Scientific American, El Mundo, japanese TV), allowing millions of persons to discover aspects of the activities of Inria. It was featured in the report of “Stratégie Nationale de Recherche France 2020”. Leveraging this visibility, the trademark “Poppy” was protected in 10 countries and its exploitation forms a pillar of the economic exploitation of Poppy: a first step was a contractualization with the company Generation Robots, who is now a distributor of Poppy kits, and a second step in preparation is the creation of a startup company. The platform is now being used as an educational platform in several engineering schools (e.g. Ensam Bordeaux and Paris, Enseirb, ENST Brest), FabLabs and science museums in Paris, Bordeaux, Rennes, Grenoble, Caen, Toulouse, and several towns in Spain, UK, Germany and the US. It has also been used in several artistic projects and associated public performances. The platform was selected as finalist for the Global Fab Award 2014, (https://www.fab10.org/en/awards ) which select the best worldwide projects in the Makers ecosystem.
Web: http://flowers.inria.fr , http://www.pyoudeyer.com and http://www.poppy-project.org
This research program, framed in the field of Developmental Robotics, focuses on a selected set of developmental mechanisms known to be key in human development (intrinsic motivation, maturation, imitation, joint attention, properties of embodiment), with the primary goal of studying how they can be functionally formalized and transposed from humans to robots and explore how they can allow robots to be equipped with important aspects of developmental capabilities observed in humans. Besides this fundamental question, the combination of these techniques with classical robot engineering techniques, and their application to the new field of personal robotics, where robots need to adapt to non-expert users in everyday environments, is studied. A secondary goal of this research program is to use those computer and robotic models as tools for understanding particular aspects of human development.
Several key mechanisms allowing robot life-long and autonomous learning of novel sensorimotor and social skills were elaborated and experimented.
First, mechanisms of active curiosity-driven exploration, inspired by human intrinsic motivation systems, were devised to allow a robot to efficiently discover how to use its initially unknown body, and the interaction with external physical objects. This includes learning skills such as omnidirectional legged locomotion, arm inverse kinematics, or controlling external flexible objects such as a fishing rod.
Second, mechanisms inspired by human maturation were elaborated, allowing the robot to adaptively freeze and free novel degrees of freedom in both motor and sensori channels. This was shown to allow exploration and learning in large unbounded sensorimotor spaces, and to considerably speed up active learning techniques.
Third, it was explored for the first time how open-ended curiosity-driven learning of multiple sensorimotor tasks could be combined with imitation learning. These mechanisms allow robot to both set their own goals and try autonomous learning to solve them, and at the same time to benefit from human advice towards the interest of certain goals or to benefit from examples of strategies to achieve them.
Open-source 3D printed Poppt robotic platform for open science and education. The Poppy platform, developed within the ERC project, is the first complete open-source 3D printed humanoid platform in the world: http://www.poppy-project.org
The Poppy platform was initially designed with the goal of developing an experimental platform allowing the systematic scientific study of the role of morphology in sensorimotor skill acquisition (morphology can be changed easily and fastly). This has been a major innovation: this was before impossible because robot platforms were developed using classical machining techniques requiring a lot of time, energy and funding. Several research labs in Europe have already began to use the Poppy platform for their own project.
As all aspects of the platform were designed to be highly modular, modifiable, robust, easily replicable, cheap, and accessible to beginners, this allowed us to apply and initiate its transfer towards education in lycées, engineering schools and universities, but also FabLabs, science museums and art/science projects. It is now becoming a major pedagogical innovation, targeting education to the digital world and its interaction with the physical world through integration of the robotic platform with software web tools to support the community and with interdisciplinary pedagogical content co-designed with users. No direct competitors exist so far.
Poppy has been presented in highly visible and prestigious wide audience venues (François Hollande at Elysée, French Sénat, Le Web conference, Tedx Cannes), in numerous high quality media/press articles (e.g. Le Monde, Scientific American, El Mundo, japanese TV), allowing millions of persons to discover aspects of the activities of Inria. It was featured in the report of “Stratégie Nationale de Recherche France 2020”. Leveraging this visibility, the trademark “Poppy” was protected in 10 countries and its exploitation forms a pillar of the economic exploitation of Poppy: a first step was a contractualization with the company Generation Robots, who is now a distributor of Poppy kits, and a second step in preparation is the creation of a startup company. The platform is now being used as an educational platform in several engineering schools (e.g. Ensam Bordeaux and Paris, Enseirb, ENST Brest), FabLabs and science museums in Paris, Bordeaux, Rennes, Grenoble, Caen, Toulouse, and several towns in Spain, UK, Germany and the US. It has also been used in several artistic projects and associated public performances. The platform was selected as finalist for the Global Fab Award 2014, (https://www.fab10.org/en/awards ) which select the best worldwide projects in the Makers ecosystem.
Web: http://flowers.inria.fr , http://www.pyoudeyer.com and http://www.poppy-project.org