This project fosters general research on multi-agent robotic systems, aiming at introducing methodologies for task planning, task allocation and teamwork supervision/coordination, driven by results from Distributed AI, Hybrid Systems and Discrete Event Systems theory. Its current case study is on Soccer Robots, with regular participations in RoboCup.
The FCT project started in 2003, and ended in this period, with two-folded contributions:
Technological, because the new robots were professionally designed and manufactured, based on the past experience on RoboCup competitions of the leading institution (ISR/IST), as well as on the know-how and competence of the sub-contracted institutions (IdMind and ServoLog).
Scientific, since some of the most recent research results of the ISocRob team (ISR/IST RoboCup MSL team) have been based on the expected availability of the new omnidirectional robots.
The option to involve Portuguese companies in the development of the robots was also successful, as new robotic prototypes, and relevant know-how were created in Portugal, instead of the simpler (but costly) solution of procuring existing solutions (not fitting the specifications of our design) made abroad. Although the main goal of this proposal was the development of robots specially designed to play soccer, the expected experience gathered from this project can now be re-used by the involved companies to build new commercially robotic solutions for other purposes rather than soccer, with financial and visibility side effects for ISR/IST. Furthermore, the platforms have open hardware and software architectures, and can be sued by other groups at ISR/IST for general research on mobile robotics.
The new robots have most of their processing power concentrated on a NEC FS900 laptop, with an INTEL Centrino 1.6 GHz processor, with 512MB RAM and a 30GB disk. The laptop includes a CD-ROM, wireless 802.11b, 3 USB 2.0 ports, and 1 mini-firewire port, as well as a spare Li-Ion battery for extra autonomy.
Each robot is endowed with the following sensors:
1 AVT Marlin F-033C firewire camera
The camera is part of an omnidirectional catadioptric vision sensor, similar to the one used in the old robots.
16 sonars (SRF04 RangeFinder) disposed in a ring around the robot.
1 500 CPR encoder per motor for motor control and odometry.
1 AnalogDevices rate-gyro XRS300EB to improve orientation determination.
1 Creative Notebook Optical Mouse (800 dpi resolution and maximum speed of 1m/s) to improve position determination.
2 Sharp infrared sensors, to measure the kicker piston displacement and to detect the ball when it is between the robot fingers.
To power the electronics and motors, 2 packs of 9Ah NiMH batteries per robot are used, with a very good autonomy/weight ratio. IdMind has developed special chargers for these batteries, which allow charging the robots in maximum 3 hours, with the batteries in place, as well as running the batteries from DC current with a cable.
Plug-and-play” connections of most peripherals to the laptop where decision-making, guidance and navigation algorithms are running were used.
Other contribution in this period focused on developing a dynamic programming algorithm to solve a class of Stochastic Games called two-person zero-sum games, inspired by the reinforcement learning algorithm Minimax-Q. In each state of the game, linear programming is used to find a Nash equilibrium, which ensures optimality in a worst-case scenario. The method is then applied to a behavioral model of a robotic soccer game. The goal is to find the worst case scenario strategy for such a team, so that a lower bound for the teams performance is guaranteed.
One further tackled topic was a general formulation of relational behaviours for cooperative real robots and an example of its implementation using the pass between soccer robots of the Middle-Sized League of RoboCup. The formulation is based on the Joint Commitment Theory and the pass implementation is supported by past work on soccer robots navigation.