In many real-world problems, a complex task is typically composed of a set of subtasks that follow a certain execution order. Traditional multi-agent reinforcement learning methods perform poorly in such multi-task cases, as they consider the whole problem as one task. For such multi-agent multi-task problems, heterogeneous relationships i.e., subtask-subtask, agent-agent, and subtask-agent, are important characters which should be explored to facilitate the learning performance. This paper proposes a dynamic heterogeneous graph based agent allocation-action learning framework. Specifically, a dynamic heterogeneous graph model is firstly designed to characterize the variation of heterogeneous relationships with the time going on. Then a multi-subgraph partition method is invented to extract features of heterogeneous graphs. Leveraging the extracted features, a hierarchical framework is designed to learn the dynamic allocation of agents among subtasks, as well as cooperative behaviors. Experimental results demonstrate that our framework outperforms recent representative methods on two challenging tasks, i.e., SAVETHECITY and Google Research Football full game.

To the best of our knowledge, there have been very few work on computational algorithms for the core or its variants in MC-nets games. One exception is the work by [Hirayama, et.al., 2014], where a constraint generation algorithm has been proposed to compute a payoff vector belonging to the least core. In this paper, we generalize this algorithm into the one for finding a payoff vector belonging to ϵ-core with pre-specified bound guarantee. The underlying idea behind this algorithm is basically the same as the previous one, but one key contribution is to give a clearer view on the pricing problem leading to the development of our new general algorithm. We showed that this new algorithm was correct and never be trapped in an infinite loop. Furthermore, we empirically demonstrated that this algorithm really presented a trade-off between solution quality and computational costs on some benchmark instances.

Cognitive radio network (CRN) provides an effective way of improving efficiency and flexibility in spectrum usage. Due to the coexistence of secondary user (SU) and primary user (PU), managing interference is a critical issue to be addressed if we are to reap the full benefits. In this paper, we consider the problem of joint interference management and resource allocation in a multi-channel ad hoc CRN. We formulate the problem as an overlapping coalition formation game to maximize the sum rate of SU links while guaranteeing the quality of service (QoS) of PU links. In the game, each SU link can make an autonomous decision and is allowed to participate in one or more cooperative coalitions simultaneously to maximize its payoff. To obtain the solution of the formulated game, a distributed, self-organizing algorithm is proposed for performing coalition formation. We analyze the properties of the algorithm and show that SU links can cooperate to reach a final stable coalition structure. Compared with existing approaches, the proposed scheme achieves appreciable performance improvement in terms of the sum rate of SU links, which is demonstrated by simulation results.

This study tries to construct an accurate ranking method for five team ball games at the Olympic Games. First, the study uses a statistical rating method for team ball games. A single parameter, called a rating, shows the strength and skill of each team. We assume that the difference between the rating values explains the scoring ratio in a match based on a logistic regression model. The rating values are estimated from the scores of major international competitions that are held before the Rio Olympic Games. The predictions at the Rio Olympic Games demonstrate that the proposed method can more accurately predict the match results than the official world rankings or world ranking points. The proposed method enabled 262 correct predictions out of 370 matches, whereas using the official world rankings resulted in only 238 correct predictions. This result shows a significant difference between the two criteria.

Card-based cryptographic protocols provide secure multi-party computations using a deck of physical cards. The most important primitive of those protocols is the shuffling operation, and most of the existing protocols rely on uniform cyclic shuffles (such as the random cut and random bisection cut) in which each possible outcome is equally likely and all possible outcomes constitute a cyclic subgroup. However, a couple of protocols with non-uniform and/or non-cyclic shuffles were proposed by Koch, Walzer, and Härtel at Asiacrypt 2015. Compared to the previous protocols, their protocols require fewer cards to securely produce a hidden AND value, although to implement of such unconventional shuffles appearing in their protocols remains an open problem. This paper introduces “pile-shifting scramble,” which can be a secure implementation of those shuffles. To implement such unconventional shuffles, we utilize physical cases that can store piles of cards, such as boxes and envelopes. Therefore, humans are able to perform the shuffles using these everyday objects. Furthermore, we show that a certain class of non-uniform and/or non-cyclic shuffles having two possible outcomes can be implemented by the pile-shifting scramble. This also implies that we can improve upon the known COPY protocol using three card cases so that the number of cases required can be reduced to two.

In this paper, we study a control problem of a concurrent discrete event system, where several subsystems are partially synchronized via shared events, under local and global constraints described by linear temporal logic formulas. We propose a hierarchical control architecture consisting of local supervisors and a coordinator. While the supervisors ensure the local requirements, the coordinator decides which shared events to be disabled so as to satisfy the global specification. First, we construct Rabin games to obtain local supervisors. Next, we reduce them based on shared transitions. Finally, we construct a global Rabin game from the reduced supervisors and a deterministic Rabin automaton that accepts every run satisfying the global specification. By solving it, we obtain a coordinator that disables shared events to guarantee the global requirement. Moreover, the concurrent system controlled by the coordinator and the local supervisors is deadlock-free.

This paper investigates open-loop Stackelberg games for a class of stochastic systems with multiple players. First, the necessary conditions for the existence of an open-loop Stackelberg strategy set are established using the stochastic maximum principle. Such conditions can be represented as solvability conditions for cross-coupled forward-backward stochastic differential equations (CFBSDEs). Second, in order to obtain the open-loop strategy set, a computational algorithm based on a four-step scheme is developed. A numerical example is then demonstrated to show the validity of the proposed method.

Card-based protocols enable us to easily perform cryptographic tasks such as secure multiparty computation using a deck of physical cards. Since the first card-based protocol appeared in 1989, many protocols have been designed. A protocol is usually described with a series of somewhat intuitive and verbal descriptions, such as “turn over this card,” “shuffle these two cards,” “apply a random cut to these five cards,” and so on. On the other hand, a formal computational model of card-based protocols via abstract machine was constructed in 2014. By virtue of the formalization, card-based protocols can be treated more rigorously; for example, it enables one to discuss the lower bounds on the number of cards required for secure computations. In this paper, an overview of the computational model with its applications to designing protocols and a survey of the recent progress in card-based protocols are presented.

In this paper, an infinite-horizon team-optimal incentive Stackelberg strategy is investigated for a class of stochastic linear systems with many non-cooperative leaders and one follower. An incentive structure is adopted which allows for the leader's team-optimal Nash solution. It is shown that the incentive strategy set can be obtained by solving the cross-coupled stochastic algebraic Riccati equations (CCSAREs). In order to demonstrate the effectiveness of the proposed strategy, a numerical example is solved.

In the obnoxious facility game with a set of agents in a space, we wish to design a mechanism, a decision-making procedure that determines a location of an undesirable facility based on locations reported by the agents, where we do not know whether the location reported by an agent is where exactly the agent exists in the space. For a location of the facility, the benefit of each agent is defined to be the distance from the location of the facility to where the agent exists. Given a mechanism, all agents are informed of how the mechanism utilizes locations reported by the agents to determine a location of the facility before they report their locations. Some agent may try to manipulate the decision of the facility location by strategically misreporting her location. As a fair decision-making, mechanisms should be designed so that no particular group of agents can get a larger benefit by misreporting their locations. A mechanism is called group strategy-proof if no subset of agents can form a group such that every member of the group can increase her benefit by misreporting her location jointly with the rest of the group. For a given mechanism, a point in the space is called a candidate if it can be output as the location of the facility by the mechanism for some set of locations reported by agents. In this paper, we consider the case where a given space is a tree metric, and characterize the group strategy-proof mechanisms in terms of distribution of all candidates in the tree metric. We prove that there exists a group strategy-proof mechanism in the tree metric if and only if the tree has a point to which every candidate has the same distance.

Assume that Alice, Bob, and Carol, each of whom privately holds a one-bit input, want to learn the output of some Boolean function, say the majority function, of their inputs without revealing more of their own secret inputs than necessary. In this paper, we show that such a secure three-input function evaluation can be performed with a deck of real cards; specifically, the three players can learn only the output of the function using eight physical cards — four black and four red cards — with identical backs.

Coalition Structure Generation (CSG) is a main research issue in the domain of coalition games. A majority of existing works assume that the value of a coalition is independent of others in the coalition structure. Recently, there has been interest in a more realistic settings, where the value of a coalition is affected by the formation of other coalitions. This effect is known as externality. The focus of this paper is to make use of Maximum Satisfiability (MaxSAT) to solve the CSG problem where externalities may exist. In order to reduce the exponentially growing number of possible solutions in the CSG problem, we follow the previous works by representing the CSG problem as sets of rules in MC-nets (without externalities) and embedded MC-nets (with externalities). Specifically, enlightened by the previous MC-net-based algorithms exploiting the constraints among rule relations to solve the CSG problem, we encode such constraints into weighted partial MaxSAT (WPM) formulas. Experimental results demonstrate that an off-the-shelf MaxSAT solver achieves significant improvements compared to the previous algorithm for the same set of problem instances.

Coalition Structure Generation (CSG) means partitioning agents into exhaustive and disjoint coalitions so that the sum of values of all the coalitions is maximized. Solving this problem could be facilitated by employing some compact representation schemes, such as marginal contribution network (MC-net). In MC-net, the CSG problem is represented by a set of rules where each rule is associated with a real-valued weights, and the goal is to maximize the sum of weights of rules under some constraints. This naturally leads to a combinatorial optimization problem that could be solved with weighted partial MaxSAT (WPM). In general, WPM deals with only positive weights while the weights involved in a CSG problem could be either positive or negative. With this in mind, in this paper, we propose an extension of WPM to handle negative weights and take advantage of the extended WPM to solve the MC-net-based CSG problem. Specifically, we encode the relations between each pair of agents and reform the MC-net as a set of Boolean formulas. Thus, the CSG problem is encoded as an optimization problem for WPM solvers. Furthermore, we apply this agent relation-based WPM with minor revision to solve the extended CSG problem where the value of a coalition is affected by the formation of other coalitions, a coalition known as externality. Experiments demonstrate that, compared to the previous encoding, our proposed method speeds up the process of solving the CSG problem significantly, as it generates fewer number of Boolean variables and clauses that need to be examined by WPM solver.

Dynamic spectrum access is the key approach in cognitive wireless regional area networks, and it is adopted by secondary users to access the licensed radio spectrum opportunistically. In order to realize real-time secondary spectrum usage, a dynamic spectrum access model based on stochastic differential games is proposed to realize dynamic spectrum allocation; a Nash equilibrium solution to the model is given and analyzed in this paper. From an overall perspective, the relationships between available spectrum percentage and the spectrum access rate are studied. Changes in the available spectrum percentage of the cognitive wireless regional area networks involve a deterministic component and a stochastic component which depends upon an r-dimensional Wiener process. The Wiener process represents an accumulation of random influences over the interval, and it reflects stochastic and time-varying properties of the available spectrum percentage. Simulation results show that the dynamic spectrum access model is efficient, and it reflects the time-varying radio frequency environment. Differential games are useful tools for the spectrum access and management in the time-varying radio environment.

Evolutionary stability has been discussed as a fundamental issue in single-criterion games. We extend evolutionarily and neutrally stable strategies to multicriteria games. Keeping in mind the fact that a payoff is given by a vector in multicriteria games, we provide several concepts which are coincident in single-criterion games based on partial vector orders of payoff vectors. We also investigate the hierarchical structure of our proposed evolutionarily and neutrally stable strategies. Shapley had introduced concepts such as strong and weak equilibria. We discuss the relationship between these equilibria and our proposed evolutionary stability.

This paper presents the development of a sound–specific vibration interface and its evaluation results by playing three commercial games with the interface. The proposed interface complements the pitfalls of existing frequency–based vibration interfaces such as vibrating headsets, mouses, and joysticks. Those interfaces may bring negative user experiences by generating incessant vibrations because they vibrate in response to certain sound frequencies. But the proposed interface which responds to only target sounds can improve user experiences effectively. The hardware and software parts of the interface are described; the structure and the implementation of a wrist pad that delivers vibration are discussed. Furthermore, we explain a sound-matching algorithm that extracts sound characteristics and a GUI-based pattern editor that helps users to design vibration patterns. The results from evaluating the performance show that the success ratio of the sound matching is over 90% at the volume of 20 dB and the delay time is around 400 msec. In the survey about user experiences, the users evaluates that the interface is more than four times effective in improving the reality of game playing than without using the vibration interfaces, and two times than the frequency–based ones.

This paper addresses an application of the potential game theory to a power-aware mobile sensor coverage problem where each sensor tries to maximize a probability of target detection in a convex mission space. The probability of target detection depends on a sensing voltage of each mobile sensor as well as its current position. While a higher sensing voltage improves the target detection probability, this requires more power consumption. In this paper, we assume that mobile sensors have different sensing capabilities of detecting a target and they can adaptively change sensing areas by adjusting their sensing voltages. We consider an objective function to evaluate a trade-off between improving the target detection probability and reducing total power consumption of all sensors. We represent a sensing voltage and a position of each mobile sensor using a barycentric coordinate over an extended strategy space. Then, the sensor coverage problem can be formulated as a potential game where the power-aware objective function and the barycentric coordinates correspond to a potential function and players' mixed strategies, respectively. It is known that all local maximizers of a potential function in a potential game are equilibria of replicator dynamics. Based on this property of potential games, we propose decentralized control for the power-aware sensor coverage problem such that each mobile sensor finds a locally optimal position and sensing voltage by updating its barycentric coordinate using replicator dynamics.

Model-based reinforcement learning uses the gathered information, during each experience, more efficiently than model-free reinforcement learning. This is especially interesting in multiagent systems, since a large number of experiences are necessary to achieve a good performance. In this paper, model-based reinforcement learning is developed for a group of self-interested agents with sequential action selection based on traditional prioritized sweeping. Every single situation of decision making in this learning process, called extensive Markov game, is modeled as n-person general-sum extensive form game with perfect information. A modified version of backward induction is proposed for action selection, which adjusts the tradeoff between selecting subgame perfect equilibrium points, as the optimal joint actions, and learning new joint actions. The algorithm is proved to be convergent and discussed based on the new results on the convergence of the traditional prioritized sweeping.

The first part of this paper contains an introduction to Bell inequalities and Tsirelson's theorem for the non-specialist. The next part gives an explicit optimum construction for the "hard" part of Tsirelson's theorem. In the final part we describe how upper bounds on the maximal quantum violation of Bell inequalities can be obtained by an extension of Tsirelson's theorem, and survey very recent results on how exact bounds may be obtained by solving an infinite series of semidefinite programs.

Quantum random access coding (QRAC) is one of the basic tools in quantum computing. It uses a quantum state for encoding the sender's bit string so that the receiver can recover any single bit of the bit string with high probability. This article surveys recent developments of QRAC, with some concrete examples of QRAC using one quantum bit, and its applications, focusing on communication complexity and locally decodable codes.