The simulator, acting as a server, accepts action commands from fully distributed clients (agents) throughout a 100-millisecond cycle and then updates the world state all at once at the end of the cycle. Agents receive sensory perceptions from the simulator asynchronously and at unpredictable intervals. We view robotic soccer as an example of a periodic team synchronization (PTS) domain. We define PTS domains as domains with the following characteristics: There is a team of autonomous agents A that collaborate toward the achievement of a joint long-term goal G. Periodically, the team can synchronize with no restrictions on communication: The agents can in effect inform each other of their entire internal states and decision-making mechanisms with no adverse effects on the achievement of G. These periods of full communication can be thought of as times at which the team is "offline."
Individual agent skills, such as kicking and dribbling (running with the ball), are important prerequisites for team collaboration. For each of these skills, many parameters affect the details of the skill execution. For example, in the ball skill of dribbling, there are parameters that affect how quickly the agent runs, how far ahead it kicks the ball, and on which side of its body the agent keeps the ball while it dribbles. The settings for these parameters usually involve a tradeoff, such as speed versus safety or power versus accuracy. It is important to gain an understanding of what exactly these tradeoffs are before "correct" parameter settings can be made. We created a trainer client that connects to the server as an omniscient offline coach client.
The CMUNITED-99 simulator team became the 1999 RoboCup simulator league champion by winning all 8 of its games, outscoring opponents by a combined score of 110-0. CMUNITED-99 builds on the successful CMUNITED-98 implementation but also improves on it in many ways. This article gives an overview of CMUNITED-99's improvements over CMUNITED-98.
The algorithm successfully detects and tracks 11 objects (5 teammates, 5 opponents, and 1 ball) at 30 frames a second. The algorithm determines the position and orientation for the robots. In addition, a Kalman-Bucy filter (Kalman and Bucy 1961) is used as a predictor of the ball's trajectory. This prediction is an integral factor in our robots' control and strategic decisions. Before developing strategic behaviors, the robots need a general control mechanism.
The CMUNITED small-robot team became the 1998 RoboCup small-robot league champion, repeating its 1997 victory. cmunited-98 built on the success of cmunited-97 and involved a number of improvements. This article gives an overview of the cmunited-98 team, focusing on this year's improvements. It concludes with the results of the RoboCup-98 competition.