MRL 2016 - Special Issue on Multiobjective Reinforcement Learning: Theory and Applications

Many real-life problems involve dealing with multiple objectives. For example, in network routing the criteria consist of energy consumption, latency, and channel capacity, which are in essence conflicting objectives. When system designers want to optimize more than one objective, it is not always clear a priori which objectives are correlated and how they influence each other upon inspecting the problem at hand. As sometimes objectives are conflicting, there usually exists no single optimal solution. In those cases, it is desirable to obtain a set of trade-off solutions between the objectives.
This problem has in the last decade also gained the attention of many researchers in the field of Reinforcement Learning (RL). RL addresses sequential decision problems in initially (possibly) unknown stochastic environments. The goal is the maximization of the agent's reward in a potentially unknown environment that is not always completely observable. Until now, there has been no special issue in a journal or a book on reinforcement learning that covered the topic of multiobjective reinforcement learning.
State of the art
We consider the extension of RL to multiobjective (stochastic) rewards (also called utilities in decision theory). Techniques from multi-objective optimization are often used for multi-objective RL in order to improve the exploration-exploitation tradeoff. Multi-objective optimization (MOO), which is a sub-area of multi-criteria decision making (MCDM), considers the optimization of more than one objective simultaneously and a decision maker decides either which solutions are important for the user or when to present these solutions to the user for further consideration. Currently, MOO algorithms are seldom used for stochastic optimization, which makes it an unexplored but very promising research area. The resulting algorithms are a hybrid between MCDM and stochastic optimization. The RL algorithms are enriched with the intuition and computational efficiency of MOO in handing multi-objective problems.
Aim and scope
The main goal of this special issue is to solicit research on multi-objective reinforcement learning. We encourage submissions
? describing applications of MO methods in RL with a focus on optimization in difficult environments that are possibly dynamic, uncertain and partially observable.
? offering theoretical insights in online or offline learning approaches for multi-objective problem domains.
Topics of interests
We enthusiastically solicit papers on relevant topics such as:
? Reinforcement learning algorithms for solving multi-objective sequential decision making problems
? Dynamic programming techniques and adaptive dynamic programming techniques handling multiple objectives
? Theoretical results on the learnability of optimal policies, convergence of algorithms in qualitative settings, etc.
? Decision making in dynamic and uncertain multi-objective environments
? Applications and benchmark problems for multi-objective reinforcement learning.
? Novel frameworks for multi-objective reinforcement learning
? Real-world applications in engineering, business, computer science, biological sciences, scientific computation, etc. in dynamic and uncertain environments solved with multi-objective reinforcement learning
Important dates
? Submissions open: December 1st 2015
? Submissions close: March 5th 2016
? Notification of acceptance: May 15th 2016
? Final manuscript due: 1 August 2016
? Expected publication date (online): November 2016