Methodology

The increased availability of affordable robot arms and grippers in the last decade has opened new perspectives for various applications. However, such robotic manipulation systems are unreliable because of missing intelligence, which leads to lack of capabilities of interpreting working and environmental conditions. A key challenge of intelligent robotics is to create robots that are capable to directly and autonomously interact with and manipulate the world around them to achieve their goals. Learning is essential to such systems, as the real world contains too many variations for a robot to have an accurate model of human requests and behaviour, of the surrounding environment, the objects in it, or the skills required to manipulate them, in advance.

For autonomous manipulation systems to be of practical interest for real world applications, the learning process cannot be started from the scratch for each new problem and system, it must be as short as possible in order to make it affordable and safe, leveraging on a priori knowledge and models about the environment and tasks. Existing AI-algorithms for manipulation are not affordable in practical applications because of the need of long training stages due to the limited results in reusing experience. It is therefore of crucial importance to be able to transfer knowledge and abilities from an application to another and from a robotic system to another.

A further aspect is the safety of the system that must be guaranteed at any time. A fundamental requirement to enable the use of autonomous robots is to detect conditions in which they are not able to solve the problem respecting the safety requirements. To achieve this goal, environment interpretation and human interaction for continuous learning up to reach a suitable “confidence level” at both the robot and the human side are required.

The EU-funded research and innovation project IntelliMan addresses these challenges by developing a novel AI-Powered Manipulation System with persistent learning capabilities, able to perceive the main characteristics and features of its surrounding by means of a heterogeneous set of sensors, able to decide how to execute a task in an autonomous way and able to detect failures in the task execution in order to request new knowledge through the interaction with humans and the environment.

IntelliMan’s main goal is to enable robots to efficiently learn how to interact with and manipulate their surroundings in a purposeful and highly performant way. IntelliMan will range from learning individual manipulation skills from human demonstration, to learning abstract descriptions of a manipulation task suitable for high-level planning, and to discovering an object’s functionality by interacting with it, in order to guarantee performance and safety.

IntelliMan has set the following specific objectives that will be reached by a cohesive action plan guided by an integrated management approach.

The project has a wide variety of specialized deployment scenarios of manipulation robots in unstructured environments that neither it nor its designers have foreseen or encountered before. The potential for autonomous manipulation application with robots capable of manipulation their environment is enormous: hospitals, elder- and child-care, factories, outer space, restaurants, service industries, home environment.

The expected outcomes of the project can be summarized in:

• Broader adoption of AI-oriented methods in robotic manipulation and human-machine interaction to solve issues related to portability, affordability, reliability and safety
• Integration of AI algorithms in prosthesis and service robots to improve acceptability and reliability
• Penetration in industry through safe, fast and easily adaptable AI-powered manipulation systems
• Rising awareness about the benefits of AI-oriented manipulation methods
• Better acceptance of robots by the general population.