Decentralised decision making in heterogeneous teams using anonymous optimisation

Mathews, George; Durrant‐Whyte, Hugh; Prokopenko, Mikhail

doi:10.1016/j.robot.2008.10.020

Cited by 33 publications

(22 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Of this family, the distributed gradient-descent based optimization (DGO) algorithm is guaranteed to find an optimal solution to quadratic optimization problems (Tsitsiklis et al, 1986;Mathews et al, 2009). The DGO algorithm is a cooperative distributed problem solving method.…”

Section: Solving Dsucs With No Constraintsmentioning

confidence: 99%

Distributed economic dispatch of embedded generation in smart grids

Dibangoye

Doniec

Fakham

et al. 2015

Engineering Applications of Artificial Intelligence

View full text Add to dashboard Cite

Section: Solving Dsucs With No Constraintsmentioning

confidence: 99%

Distributed economic dispatch of embedded generation in smart grids

Dibangoye

Doniec

Fakham

et al. 2015

Engineering Applications of Artificial Intelligence

View full text Add to dashboard Cite

“…consensus, motion planning) or discrete (ie. TA problems such as mission planning and vehicle routing) [2]. The focus in this work is on the TA problem.…”

Section: Previous Workmentioning

confidence: 99%

Mission planning in unstructured environments: A reinforcement learning approach

Basso

Durrant‐Whyte

Hedrick

2011

IEEE Conference on Decision and Control and European Control Conference

Self Cite

View full text Add to dashboard Cite

Mission planning scenarios in robotics typically involve one or more semi-autonomous agents and a human operator. The high-level goal is to find an optimal allocation of agents to tasks. Each individual task or collection of tasks may have subgoals, such as search, localization, or information gathering. Each of these lower-level goals are their own topics, and the high-level goal of task accomplishment can be categorized as a decision problem. The focus of this work is on solving decision problems through reinforcement learning. Tasks are completely parameterized by a minimal set of basis constraints-spatial, temporal, and (agent-task) coupling-which produces a single cost for each agent-task pairing. The cost completely captures the ability of a specific agent to perform a specific task. A novel state representation is presented that mitigates exponential growth of the state space by scaling independently of the spatial dimension. The decision problem is cast as an MDP and an optimal policy is found using Q-learning. Simulation results are presented for a two-agent two-task example, showing convergence of the value function and improved learning over time. To highlight the scalability of the learning algorithm, additional simulations compare the learned policy to a handcoded greedy policy for varying number of tasks. The learned policy is shown to reliably allocate agents to tasks with minimal parameter tuning and is robust to low-level changes to agent dynamics as well as random agent motion.

show abstract

“…The most substantial difference is that, often, global system goals are fixed by the task statement, in other words, the task is assigned to the global system and the component control systems goals must be crafted so that when all the components are operating together this global goal is achieved. Hierarchical or sequential task/goal decompositions may be found, so that the local optimization performed by each agent contributes to the global optimum seeking process [104]. This is often achieved in terms of different utility based strategies derived from the agent literature that permit defining private utility functions that will lead to the desired global utility under the appropriate environmental conditions.…”

Section: Controlmentioning

confidence: 99%

“…In the latter, the fulfilment of the global goal emerges spontaneously (synergistically) from the independent fulfilment of the individual tasks/goals. Evidently, the individual tasks must have been formulated in an appropriate way and the global problem must be decomposable (a common research issue in agent literature) [51,104]. This decomposition may be achieved by negotiation as in [130,145], by some form of swarm dynamics [190], or even by some form of learning [17].…”

Section: Controlmentioning

confidence: 99%