Prasanna Lokuge

Improving the adaptability in automated vessel scheduling in container ports using intelligent software agents

Abstract Faster turnaround time of vessels and high berth productivity are paramount factors in container terminals for assuring competitive advantage in the shipping industry. An autonomous decision-making capability in the terminal is vital in achieving the required productivity. Vessel scheduling/berthing system in a container terminal is regarded as a very complex dynamic application in today’s business world. The Artificial Intelligence (AI) community has been researching in the field of intelligent (or rational) agents for more than a decade and implementations are found in many commercial applications. The Beliefs, Desires and Intention (BDI) agent architecture is probably the most mature model for many industrial applications in today’s context. However, it is not the best agent model for complex applications that must learn and adapt their behaviours in making rational decisions. We propose a new hybrid BDI framework with an intelligent module to overcome the limitations in the generic BDI model. Learning and the adaptability of the environments are assured with the introduction of the Knowledge Acquisition Module (KAM) in the generic BDI architecture in our proposed framework. The dynamic selection of the intention structures has been improved with a trained neural network. The knowledge required to handle vagueness or uncertainty in the environment has been modelled with an Adaptive Neuro Fuzzy Inference System (ANFIS) in berths. Finally, the benefits and the usability of hybrid BDI model for a vessel berthing application is discussed with experiential results.

Hybrid BDI agents with improved learning capabilities for adaptive planning in a container terminal application

Vessel berthing system in a container terminal is regarded as a very complex dynamic application in todays business world. We propose a new extended BDI framework with an intelligent module for handling complex situations. Change rate of beliefs (/spl Psi/) and expected cost of reaching the final goal state from different states in the plan hierarchy have been considered by the agent in the proposed architecture. This would enable agents to identify the alternative plans in the intention structure with the change of the environment. Dynamic selection of plans and expected cost of achieving the final goal state from various plan paths are modeled with the use of a supervised neural network. Adaptive neuro fuzzy inference system (ANFIS) has been incorporated in making the final rational decisions in the agent model.

Collaborative neuro-BDI agents in container terminals

Berth scheduling and monitoring of the vessel operations are of paramount importance in order to assure faster turnaround time and high productivity of any container terminal. The need for an intelligent system that dynamically adapts to the changing environment is apparent, as there are a limited number of berths and resources available in container terminals for delivering services to vessels. We discuss how BDI (Beliefs, Desires and Intentions) agents can be supported with Neural Network and fuzzy logic in a collaborative environment of a multi agents system for the scheduling and monitoring of vessel berths in container ports. Straightforward plans are handled by the generic BDI architecture. Complex planning which requires the learning and adaptability behavior is modeled with neural networks. Beliefs with fuzzy scenarios are modeled with fuzzy logic enabling agents to make rational decisions in the environment of uncertainty. Agents can autonomously adapt to the changing environment in assigning berths for vessels.

Reinforcement learning in neuro BDI agents for achieving agent's intentions in vessel berthing applications

Complex business application systems that involve non trivial decision making can have highly unpredictable situations. In such situation adaptive and intelligent behaviors would able to mitigate the risk in business. Vessel berthing application in container terminals is regarded as a very complex dynamic application, which requires autonomous decision making capabilities to improve the productivity of the berths. On the other hand, BDI agent systems have been implemented in many applications and found some limitations in learning. We propose a new enhanced hybrid BDI model with ANFIS and reinforcement learning methods to over come the above limitation. Our paper discusses how the commitment strategy of agents desire, intentions and plans could be enhanced with intelligent learning capabilities. A new motivation based distance calculation method supported with ANFIS and reinforcement learning is proposed in the paper, which improve the reactive, proactive and intelligent behaviors of generic BDI agents in complex applications.

BDI Agents Using Neural Network and Adaptive Neuro Fuzzy Inference for Intelligent Planning in Container Terminals

Vessel berthing operations in a container terminal is a very complex application since environmental changes should be considered in assigning a right berth for a vessel. Dynamic planning capabilities would essentially enhance the quality of the decision making process in the terminals. Limitations in the social ability and learning capabilities of the generic BDI execution cycle have been minimized in the proposed architecture. Paper describes the use of Belief-desires-intention (BDI) agents with neural network and adaptive neuro fuzzy inference system in building the intelligence especially in planning process of the agent. Previous knowledge and the uncertainty issues in the environment are modeled with the use of intelligent tools in the hybrid BDI agent model proposed in the paper. This would essentially improve the adaptability and autonomy features of the BDI agents, which assures better planning, scheduling and improved productivity of the terminal.

Hybrid BDI Agents with ANFIS in Identifying Goal Success Factor in a Container Terminal Application

Faster turnaround time of vessels and high berth productivity are paramount important factors of any container terminals in assuring competitive advantage in the shipping industry. The Paper proposes a hybrid BDI agent model with Neural Networks and Adaptive Neurofuzzy Inference system (ANFIS) in dealing with complex environments such as operations in container terminal in the shipping industry. Hybrid model is emphasized to improve the learning capabilities of the generic BDI agents. A plan Tuple: PLAN ¿ B, D, I, SF, TY ¿ is introduced in handling plans in the intention structure where, B- Beliefs, D - Desires I - intentions, ST - success factor and TY for the type of the plan. Learning capabilities and handling partially successful goal states in the BDI agent model have been improved with the introduction of the hybrid architecture suggested.

Implementing intention Reconsideration Strategies and Deliberation Decision of Intelligent h-BD[I] Agents

We present a novel framework that enables h-BD[I] agents to dynamically choose Its intention reconsideration policy in order to perform optimally in accordance with the dynamic changes in the environment. Paper discusses the present limitations of BDI (belief-desire-intention) agent model and proposes a new extended architecture, A-BD[I] for non deterministic, dynamic environments. The lack of learning competences and difficulties in dealing with vague or imprecise data sets in the environment are the main obstacles in finding an optimal solution in the present BDI model. Paper discuss the means-end reasoning of A-BD[I] agents in dynamic environments which execute plans for achieving the agents intention.

Shared learning vector quantization in a new agent architecture for intelligent deliberation

The basic belief-desire-intention (BDI) agent model appears to be inappropriate for building complex system that that must learn and adapt their behaviour dynamically. The contribution of the paper is the introduction of a new “intelligent-Deliberation” process in the hybrid BDI (h-BD[I]) architecture that enables an improved decision making features in a dynamic, and complex environment. Shared learning vector quantization (SLVQ) based neural network is proposed for the intelligent deliberation of the agent model. Paper discusses the benefits of incorporating knowledge based techniques in the deliberation process of the extended h-BD[I] model.