Register now After registration you will be able to apply for this opportunity online.
This opportunity is not published. No applications will be accepted.
Coordinated Defensive Gameplay for RoboCup
We wish to devise an optimized defensive gameplay strategy for our Nao robots as part of the Nomadz Robocup Team at ETH Zürich (https://robocup.ethz.ch/).
Building on a previous project that developed an offensive gameplay strategy for our team in the 5-a-side Standard Platform League of RoboCup, we wish to determine the coordinated defensive strategy to be implemented when our team does not have possession of the ball. This strategy should minimize the chance of conceding a goal, while considering uncertain ball position, probabilities of winning the ball in a one-on-one duel, and possible opponent strategies.
The project will build upon the current capabilities of our Nao robots, which includes vision-based self-localization, walking, and passing. Various optimization methods will be considered, including solving mixed integer problems, dynamic programming, and potentially incorporating additional sources of nonlinearities due to geometric modeling or gameplay decisions. The project should strike a balance between model fidelity and tractability of the optimization methods available to solve the problem.
The resulting strategy can take many different forms, including incorporating heuristic strategies or simpler policies, in addition to higher level optimization. Regardless, the final strategy should weigh the various options available, and be arrived at in a logical and deliberative manner.
Building on a previous project that developed an offensive gameplay strategy for our team in the 5-a-side Standard Platform League of RoboCup, we wish to determine the coordinated defensive strategy to be implemented when our team does not have possession of the ball. This strategy should minimize the chance of conceding a goal, while considering uncertain ball position, probabilities of winning the ball in a one-on-one duel, and possible opponent strategies.
The project will build upon the current capabilities of our Nao robots, which includes vision-based self-localization, walking, and passing. Various optimization methods will be considered, including solving mixed integer problems, dynamic programming, and potentially incorporating additional sources of nonlinearities due to geometric modeling or gameplay decisions. The project should strike a balance between model fidelity and tractability of the optimization methods available to solve the problem.
The resulting strategy can take many different forms, including incorporating heuristic strategies or simpler policies, in addition to higher level optimization. Regardless, the final strategy should weigh the various options available, and be arrived at in a logical and deliberative manner.
The goal of this project is to calculate defensive strategy in real-time on the Nao robots. This involves work in problem formulation, simulation, and embedded coding on the robots. The main focus will be on modeling and development of the defensive strategy, but teaching assistants will be available to help implement the result on the actual robots.
The goal of this project is to calculate defensive strategy in real-time on the Nao robots. This involves work in problem formulation, simulation, and embedded coding on the robots. The main focus will be on modeling and development of the defensive strategy, but teaching assistants will be available to help implement the result on the actual robots.
Alexandros Tanzanakis (atanzana@ethz.ch), Benjamin Flamm (flammb@control.ee.ethz.ch)
Alexandros Tanzanakis (atanzana@ethz.ch), Benjamin Flamm (flammb@control.ee.ethz.ch)