Keepaway Soccer: From Machine Learning Testbed to Benchmark

By Peter Stone, Gregory Kuhlmann, Matthew E. Taylor, Yaxin Liu. 2005

158 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Reinforcement Learning, Evaluation & Benchmarks

How researchers see this paper

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

In Artificial Intelligence

In Reinforcement Learning

In Evaluation & Benchmarks

Citations in the corpus

Benchmarking Deep Reinforcement Learning for Continuous Control (2016, methodology): “Tdlearn (Dann et al., 2014) includes a 20-link pole balancing task, DotRL (Papis & Wawrzy´nski, 2013) includes a variable-DOF octopus arm and a 6-DOF planar cheetah model, PyBrain (Schaul et al., 2010) includes a 16-DOF humanoid robot with standing and jumping tasks, RoboCup Keepaway (Stone et al., ”
Monotonic Value Function Factorisation for Deep Multi-Agent Reinforcement Learning (2020, background): “While some multi-agent testbeds have emerged, such as Poker (Heinrich and Silver, 2016), Pong (Tampuu et al., 2017), Keepaway Soccer (Stone et al., 2005), or simple gridworld-like environments (Lowe et al., 2017; Leibo et al., 2017; Yang et al., 2018; Zheng et al., 2017), there are currently no…”
The StarCraft Multi-Agent Challenge (2019, background): “Stone et al. (2005) presented Keepaway soccer, a domain built on the RoboCup soccer simulator (Kitano et al., 1997), a 2D simulation of a football environment with simpliﬁed physics, where the main task consists of keeping a ball within a pre-deﬁned area where agents in teams can reach, steal, and…”
QMIX: Monotonic Value Function Factorisation for Deep Multi-Agent Reinforcement Learning (2018, background): “, 2017), Keepaway Soccer (Stone et al., 2005), or simple gridworld-like environments (Lowe et al.”
Transfer Learning via Inter-Task Mappings for Temporal Difference Learning (2007, methodology): “All our experiments are run on a code base derived from version 0.6 of the benchmark Keepaway implementation 2 (Stone et al., 2006) and the RoboCup Soccer Server version 9.4.5.”
Batch reinforcement learning in a complex domain (2007, methodology): “Our experiments use a slightly modified version of the benchmark version of Keepaway [10, 11].”
Hierarchical Approaches for Reinforcement Learning in Parameterized Action Space (2018, background): “Tasks with this kind of action space have been proposed in the Reinforcement Learning (RL) community for a long time (Stone et al. 2006) but have not been explored much.”
Cross-domain transfer for reinforcement learning (2007, methodology): “E D U Department of Computer Sciences, The University of Texas at Austin Abstract A typical goal for transfer learning algorithms is to utilize knowledge gained in a source task to learn a target task faster.”
Integrating reinforcement learning with human demonstrations of varying ability (2011, methodology): “Experiments in this paper use the public versions 11.1.0 of the RoboCup Soccer Server [14], and 0.6 of UT-Austin’s Keepaway players [23].”
Half Field Offense in RoboCup Soccer: A Multiagent Reinforcement Learning Case Study (2006, background): “They have also been in the past used for learning keepaway [10].”
Transfer via inter-task mappings in policy search reinforcement learning (2007, background): “robot soccer Keepaway, a standard multiagent RL benchmark domain [10], and Server Job Scheduling (SJS) [17], a probabilistic autonomic computing domain.”
The Reinforcement Learning Competition 2014 (2014, background): “The 2008 competition featured Helicopter Hovering (Bagnell and Schneider 2001; Ng et al. 2006), Keepaway Soccer (Stone et al. 2006; Stone, Sutton, and Kuhlmann 2005), and Real-Time Strategy.”
The Reinforcement Learning Competitions (2010, methodology): “Keepaway (Stone et al. 2005; Stone, Sutton, and Kuhlmann 2005) is a simulated robot soccer domain built on the RoboCup Soccer Server (Noda et al. 1998), an open source software platform that has served as the basis of multiple international competitions and research challenges.”
Value Functions for RL-Based Behavior Transfer: A Comparative Study (2005, methodology): “Our CMAC and RBF agents are based on the keepaway benchmark players distributed by UT-Austin 1 which are described in (Stone et al. 2005).”
Value-Function-Based Transfer for Reinforcement Learning Using Structure Mapping (2006, background): “From (Taylor, Stone, & Liu, 2005), Keepaway is a subtask of RoboCup soccer and a publicly available benchmark for reinforcement learning (Stone et al., 2006), in which one team of n K keepers must try to keep possession of the ball within a small rectangular region, while another team of n T takers…”
Behavior transfer for value-function-based reinforcement learning (2005, background): “Our agents are based on the keepaway benchmark players distributed by UT-Austin(2) which are described in [19].”