Learning to Coordinate without Sharing Information

By Sandip Sen, Mahendra Sekaran, John Hale. 1994

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

Multiagent Learning: Basics, Challenges, and Prospects (2012, background): “Examples of MAL approaches from which this classification was derived are Haynes, Lau, and Sen (1996); Vidal and Durfee (1996) (multiplied learning), Sen, Sekaran, and Hale (1994) and Weiss (1993) (divided learning), and Bui, Kieronska, and Venkatesh (1996) and Nagendra Prasad, Lesser, and Lander…”
Sparse cooperative Q-learning (2004, methodology): “Despite the lack of guaranteed convergence, this method has been applied successfully in multiple cases (Tan, 1993; Sen et al., 1994).”
Planning, Learning and Coordination in Multiagent Decision Processes (1996, background): “For example, agents might communicate in order to determine task allocation [37, 57]; conventions (or social laws) might be imposed by the system designed so that optimal joint action is assured [31, 48]; or a coordinated policy (or conventions) might be learned through repeated interaction [47, 46,”
A Survey of Collectives (2004, background): “In such cases coordination [205], negotiations [135], coalition formation [193, 195, 249] or contracting [3] among agents may be needed to ensure that they do not work at cross purposes.”
Collaborative Multiagent Reinforcement Learning by Payoff Propagation (2006, methodology): “Despite the lack of guaranteed convergence, this method has been applied successfully in multiple cases (Tan, 1993; Sen et al., 1994).”
Predicting and Preventing Coordination Problems in Cooperative Q-learning Systems (2007, methodology): “This has been demonstrated, for example, in hexapedal robot locomotion [Maes and Brooks, 1990], network load balancing [Schaerf et al., 1995], and a cooperative boxpushing task [Sen et al., 1994].”
Multi-Agent Reinforcement Learning:a critical survey (2003, background): “The simplest way to extend this to the multi-agent SG setting is just to add a subscript to the formulation above; that is, to have the learning agent pretend that the environment is passive: Several authors have tested variations of this algorithm (e.g., [Sen et al. 1994]).”
A Survey of Learning in Multiagent Environments: Dealing with Non-Stationarity (2017, background): “However, it has been applied with some success in different multiagent scenarios (Tan, 1993; Sen et al., 1994; Crites and Barto, 1998).”
The world of independent learners is not markovian (2011, methodology): “However this approach has been successfully applied in simulation to domains such as prey-pursuing games [30], block-pushing problems [27], the control of elevators [9], the exploration of planets by mobile robot teams [41], the control of a two-link rigid manipulator [7], multi-robot cooperative tr”
A Comprehensive Survey of Multiagent Reinforcement Learning (2008, methodology): “Other taxonomy axes include the following: OAL [64] WoLF-PHC [13] PD-WoLF [74] Direct Policy Search Temporal-Difference RL Game Theory AWESOME [60] MetaStrategy [55] IGA [66] WoLF-IGA [13] GIGA [67] GIGA-WoLF [57] fictitious play [65] [69]-[71] Fig.”
The Dynamics of Reinforcement Learning in Cooperative Multiagent Systems (1998, methodology): “Tan [20] and Sen, Sekaran and Hale [16] apply RL to independent agents and demonstrate empirical convergence.”
Cooperative Multi-Agent Learning: The State of the Art (2005, result): “Similar results showing good coordination without modeling other agents are reported in [210, 209], where a couple of robots successfully learn to cooperatively push a box without either even being aware of the other’s presence.”
Multi-agent Reinforcement Learning: An Overview (2010, methodology): “Single-agent RL algorithms like Q-learning can be directly applied to the multi-agent case [105].”
Multiagent Learning of Coordination in Loosely Coupled Multiagent Systems (2015, methodology): “Despite the lack of guaranteed optimal performance, this approach has been applied successfully in certain cases [41], [42].”