Results 1 -
3 of
3
Real-Time Training of Team Soccer Behaviors
"... Abstract. Training robot or agent behaviors by example is an attractive alternative to directly coding them. However training complex behaviors can be challenging, particularly when it involves interactive behaviors involving multiple agents. We present a novel hierarchical learning from demonstrati ..."
Abstract
-
Cited by 1 (1 self)
- Add to MetaCart
(Show Context)
Abstract. Training robot or agent behaviors by example is an attractive alternative to directly coding them. However training complex behaviors can be challenging, particularly when it involves interactive behaviors involving multiple agents. We present a novel hierarchical learning from demonstration system which can be used to train both single-agent and scalable cooperative multiagent behaviors. The methodology applies manual task decomposition to break the complex training problem into simpler parts, then solves the problem by iteratively training each part. We discuss our application of this method to multiagent problems in the humanoid RoboCup competition, and apply the technique to the keepaway soccer problem in the RoboCup Soccer Simulator. 1
RoboPatriots: George Mason University 2011 RoboCup Team
"... The RoboPatriots are a team of three humanoid robots designed by the Computer Science Department at George Mason University. Each robot is based on the Kondo KHR-3HV, a customized Surveyor SVS camera, and a Gumstix ..."
Abstract
- Add to MetaCart
(Show Context)
The RoboPatriots are a team of three humanoid robots designed by the Computer Science Department at George Mason University. Each robot is based on the Kondo KHR-3HV, a customized Surveyor SVS camera, and a Gumstix
Training Heterogeneous Teams of Robots
"... Abstract. Heterogeneous multi-robot teams are common solutions to complex tasks, especially those that are inherently cooperative. Training robots, rather than coding them, to work together in these teams is an attractive prospect, but is very difficult due to the extremely large state space and the ..."
Abstract
- Add to MetaCart
(Show Context)
Abstract. Heterogeneous multi-robot teams are common solutions to complex tasks, especially those that are inherently cooperative. Training robots, rather than coding them, to work together in these teams is an attractive prospect, but is very difficult due to the extremely large state space and the inherent inverse problem which separates the agents’ micro-level behaviors and the desired macro-level emergent phenomenon. We approach this problem with HiTAB, a learning from demonstration system which uses behavior decomposition to allow rapid training of teams with minimal samples. This paper presents and compares two approaches to training teams of heterogenous robots: first, forming a multiagent control hierarchy which scales to large numbers of robots but requires the training of additional virtual controller agents; and second, modifying each robot’s feature space to include information about other robots ’ current behaviors or limited internal state. 1
