Team-partitioned, opaque-transition reinforcement learning

By Peter Stone, Manuela M. Veloso. 1999

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Citations in the corpus

Episodic Exploration for Deep Deterministic Policies: An Application to StarCraft Micromanagement Tasks (2016, background): “…policies for individual agents in a collaborative setting with communication constraints [38, 3], with applications such as soccer robot control [31], and methods such as hierarchical reinforcement learning for communicating high-level goals [12], or learning an eﬃcient communication protocol…”
Decision-Theoretic Planning: Structural Assumptions and Computational Leverage (1999, background): “Macros have a long history in classical planning and problem solving [50, 77], but only recently have they been generalized to MDPs [68, 98, 100, 109, 128, 129, 134].”
Theoretical considerations of potential-based reward shaping for multi-agent systems (2011, methodology): “One such method, empirically demonstrated to decrease the time for each individual learning in a common environment to converge on a stable policy, is incorporating heuristic knowledge [17, 25].”
Reinforcement Learning for RoboCup Soccer Keepaway (2005, methodology): “goal-scoring as the goal state, TPOT-RL ( Stone & Veloso, 1999 ) was successfully used to allow a”
A comprehensive survey on machine learning for networking: evolution, applications and research opportunities (2018, background): “Team-partitioned opaque-transition reinforcement learning (TPOT-RL), proposed by Stone and Veloso for the RoboCup-1998 (Robot Soccer World Cup II) [425], is the first fully collaborative MARL technique to be applied to packet routing [424].”
Q-Decomposition for Reinforcement Learning Agents (2003, methodology): “This update is analogous to that proposed by Stone and Veloso (1998), who assume that the action space may be partitioned into subspaces, one for each subagent, and that rewards realized by each subagent are independent given only the local action of each sub-agent.”
Multi-agent Reinforcement Learning: An Overview (2010, background): “Robot soccer is a popular, complex test-bed for MARL, that requires most of the skills enumerated above [77, 114, 115, 131, 142].”
Reinforcement learning based resource allocation in business process management (2011, methodology): “Note that an action-dependent feature [40] is used in this research.”
Hierarchical multi-agent reinforcement learning (2001, background): “There has also been work on reducing the parameters needed for Q-learning in multi-agent domains, by learning action values over a set of derived features[12].”
Computational Intelligence in Wireless Sensor Networks: A Survey (2011, background): “Learning: Team-partitioned, opaque-transition reinforcement learning (TPOT-RL) has been developed for simulated robotic soccer [49].”
Decentralized MDPs with sparse interactions (2011, background): “For example, learning tasks involving multiple agents can be partitioned in a state-wise manner, allowing different agents to independently learn the resulting " smaller tasks " [4].”
Scaling Reinforcement Learning toward RoboCup Soccer (2001, methodology): “Also using goal-scoring as the goal state, TPOT-RL ( Stone & Veloso, 1999 ) was successfully used to allow a full team of agents to learn collaborative passing and shooting policies using a Monte Carlo learning approach, as opposed to the TD learning explored in this paper.”
Layered Learning (2000, methodology): “The input representation consists of one coarse geographical component and one action-dependent feature [20] for each possible pass.”
Reinforcement Learning and Function Approximation (2005, methodology): “The idea is sketched in (Russel & Norvig 2003) and is used by (Stone & Veloso 1999) (in a setting with other complicating factors).”
From Motor Control to Team Play in Simulated Humanoid Football (2021, background): “These works typically focus on the handling of large state-action spaces (134, 135), skill learning (136–141), the keep-away and half-field offense tasks and multi-agent coordination (142–150), grounded simulation learning for improved skills (151–154) (e.”
A Comprehensive Survey of Multiagent Reinforcement Learning (2008, background): “Robot soccer is a popular, complex test-bed for MARL, that requires most of the skills enumerated above [4], [97]– [100].”