A research group at the Polytechnic University of Madrid is developing a system for coordination and communication of multi-robot systems. To do this, they use bio-inspired evolutionary algorithms and reinforcement learning for robots to adapt to the environment and learn throughout their life cycle.
Also work in the independent evolution of a language for robot teams. These skills would serve for applications in the field of transport, surveillance or disaster location. With the aim of providing solutions from the experience in other more traditional areas such as industrial robotics and autonomous, the Research Group of Computational Perception and Robotics from UPM, in collaboration with researchers.
The project focuses on the design and development of a modular architecture that provides multi-component control aspects, structural design (including kinematics and dynamics), and the mechanisms for monitoring, navigation and coordination.
Thus emerged ANLAGIS (Adaptive Neuron-like network of Granular and Learning Automata System), a control architecture proposed robotic systems using evolutionary methods and machine learning based, in particular, reinforcement learning to tune models of neurons and probabilistic automata, where the state space is granular fuzzy techniques and threshold.
These methods have been used in coordinating teams of robots for cooperative and competitive tasks in a coordinated manner (interception, movement coordination, cooperative construction of environment maps) would be very useful, for example, in transport applications of materials and ground surveillance.
The researchers also work in modular robotics line, because sometimes it is interesting that several robots, components or modules come together to create a new robot or module with a structure and shape that allows access to sites not otherwise be possible or to facilitate the completion of certain tasks.
Along these lines, have proposed methods to reduce the complexity to set standards articulated kinematic chains, and applied to the locomotion of robots serpenoides. This line is also very interesting, because it has a direct application in localization tasks, disaster monitoring in unstructured spaces.
Currently, researchers are studying the UPM emergence and evolution of language in teams of robots. The objective pursued is to prove whether a set of robots can be created independently, a language that serves to communicate with each other brands or places of an environment that were not considered at design time. For example, if a team of robots can appoint defects or areas of accumulation of garbage in their areas of surveillance.
McRobs The project has been funded by the National R + D + i and some results and conclusions of their lines of work have been published in various scientific journals such as Information Science and Neurocomputing, and also in various international conferences.
