Chi-Kong Chan

QoS-Based Cross-Layer Scheduling for Wireless Multimedia Transmissions with Adaptive Modulation and Coding

It is challenging to design effective scheduling algorithms for multimedia transmissions over wireless channels that employ adaptive modulation and coding (AMC). On the one hand, it is desirable for the overall system throughput to be enhanced by taking advantage of multi-user diversity. On the other hand, fairness or QoS guarantees need to be maintained for individual users, especially in the case of multimedia applications that have strict delay requirements. In this paper, we propose a novel scheduling algorithm called QoS-based cross-layer scheduling (QoS-CLS) to achieve a good design tradeoff. To maximize the system throughput, the algorithm takes into account the information on both the physical layer and the data link layer to schedule user transmissions. Using cross-layer information, the scheduling problem is formulated as a Markov Decision Process and the optimal decision policy (based on the channel status, traffic state, and buffer status of each traffic flow) is pre-calculated by linear programming. This policy is then stored in the system for scheduling in real-time. Simulation results show that QoS-CLS can greatly enhance the channel throughput compared to existing algorithms because of its cross-layer QoS consideration and the optimization method. Moreover, it can provide QoS guarantees while achieving efficient resource sharing among different traffic flows.

Applying genetic algorithms in database partitioning

One popular technique used to enhance database performace is attribute partitioning. Attribute partitioning is the process of subdividing the attributes of a relation and then grouping them into fragments so as to minimize the number of disk access by all transactions. On the other hand, tuple clustering, which is the process of rearranging the order of tuples so that frequently queried tuples are grouped into as few blocks as possible, is mostly ignored. In this paper, we address the need of considering the n-ary attribute partitioning and tuple clustering at the same time in a relational database. A new algorithm is proposed for mixed fragmentation design using genetic algorithm. Java programs have been developed to implement the genetic algorithm for mixed fragmentation and the results are promising. It provides an improvement over previous works which considered vertical partitioning and tuple clustering separately. Comparisons with exhaustive enumeration and random search are also presented.

An argumentation-oriented multi-agent system for automating the freight planning process

This paper advocates the use of multi-agent systems in the freight forwarding industry. We propose an intelligent mobile agent system to cope with a dynamic freight forwarding environment where up-to-date information is crucial but time-consuming to obtain. A key component of our system is an agent argumentation mechanism that allows decision support agents to discuss, argue, and come to a compromise in order to derive well-explained freight planning solutions. A number of artificial intelligence mechanisms are implemented, namely: (1) a mobile-agent-based automated information gathering mechanism, where designated mobile agents access various websites automatically to gather information (e.g., weather conditions on a candidate route) critical for cargo consolidation and route planning, (2) a fuzzy logics engine for risk evaluation, and (3) a simulated annealing engine for optimizing cargo consolidation. A system prototype is developed and the feasibility of our approach is demonstrated in a case study. A series of experiments are also conducted to evaluate the systems performance.

Multi-auction Approach for Solving Task Allocation Problem

Request for Proposal (RFP) problem is a type of task allocation problem where task managers need to recruit service provider agents to handle complex tasks composed of multiple sub-tasks, with the objective being to assign each sub-task to a capable agent while keeping the cost as low as possible. Most existing approaches either involve centralized algorithms or require each agents cost for doing each sub-task to be known publicly beforehand, or attempt to force the agents to disclose such information by means of truth-telling mechanism, which is not practical in many problems where such information is sensitive and private. In this paper, we present an efficient multi-auction based mechanism that can produce near-optimal solutions without violating the privacy of the participating agents. By including an extra verification step after each bid, we can guarantee convergence to a solution while achieving optimal results in over 97% of the times in a series of experiment.

Agent-based flight planning system for enhancing the competitiveness of the air cargo industry

In this paper, we propose a multi-agent-based framework to facilitate process automation for the air cargo industry. The focus is on enhancing two labor-intensive flight planning processes, namely cargo consolidation and equalization. By employing a software agent-based flight planning module, which is linked with an RFID-based warehouse management system, air cargo items received at a freight forwarders warehouse can be processed more efficiently and flight plans can be generated automatically. In particular, we employ agents equipped with simulated annealing optimization engines to handle the time-consuming tasks of optimization. By doing so, the latest flight plans can be generated more efficiently. The system has been evaluated experimentally by both simulated and real-life data. The results are encouraging. For example, operation steps that normally require over 30minutes to complete can now be carried out in as quickly as two minutes, and produce a better solution.

On improved least flexibility first heuristics superior for packing and stock cutting problems

Two dimensional cutting and packing problems have applications in many manufacturing and job allocation problems. In particular, in VLSI floor planning problems and stock cutting problems, many simulated annealing and genetic algorithms based methods have been proposed in the last ten years. These researches have mainly been focused on finding efficient data structures for representing packing results so the search space and processing time of the underlying search engine can be minimized. In this paper, we tackle the problem from a different approach. Instead of using stochastic searches, we introduce an effective deterministic optimization algorithm for packing and cutting. By combining an improved Least Flexibility First principle and a greedy search based evaluation routine, we can obtain very encouraging results: In stock cutting problems, our algorithm achieved over 99% average packing density for a series of public rectangle packing data sets, which is significantly better than the 96% packing density obtained by meta-heuristics (simulated annealing) based results while using much less CPU time; whereas in rectangle packing applying the well-known MCNC and GSRC benchmarks, we achieved the best (over 96%) packing density among all known published results packed by other methods. Our encouraging results seem to suggesting a new experimental direction in designing efficient deterministic heuristics for some kind of hard combinatorial problems.

Belief-based Stability in Coalition Formation with Uncertainty

Game Theoretic Analysis of Coalition Formation.- Strong Core and Weak Core.- b-Core.- An Example Application of the b-Core.- The Complete Picture-the sb-Core and the wb-Core.

Reciprocal Social Strategy in Social Repeated Games and Emergence of Social Norms

In an artificial society where agents repeatedly interact with one another, effective coordination among agents is generally a challenge. This is especially true when the participating agents are self-interested, and that there is no central authority to coordinate, and direct communication or negotiation are not possible. Recently, the problem was studied in a paper by Hao and Leung, where a new repeated game mechanism for modeling multi-agent interactions as well as a new reinforcement learning based agent learning method were proposed. In particular, the game mechanism differs from traditional repeated games in that the agents are anonymous, and the agents interact with randomly chosen opponents during each iteration. Their learning mechanism allows agents to coordinate without negotiations. The initial results had been promising. However, extended simulation also reveals that the outcomes are not stable in the long run in some cases, as the high level of cooperation is eventually not sustainable. In t...

Belief-Based Stability in Non-transferable Utility Coalition Formation

Coalition stability is an important concept in coalition formation. One common assumption in many stability criteria in non-transferable utility games is that the preference of each agent is publicly known so that a coalition is said to be stable if there is no objections by any sub-group of agents according to the publicly known preferences. However, in many applications including some software agent applications, this assumption is not true. Instead, agents are modeled as individuals with private belief and decisions are made according to those beliefs instead of common knowledge. Such belief based architectures have impacts on the coalitions stability which is not reflected in the current stability criteria. In this paper, we extend the classic stability concept of the core by proposing a new belief based stability criterion which we labeled the belief-based core.