Agent-based Formation Flight Coalition under Incomplete Information

The continued increasing air traffic demand and the corresponding fuel consumption urge the innovations of technologies and operating modes in commercial aviation community. Formation flight, due to its potential for reducing fuel use, are widely recognized as one of the most effective ways to improve aviation fuel saving. This study addresses the commercial formation coalition problem under incomplete information. First, a mathematical formulation is redefined to fit well the agent-based computation. Second, a BDI agent-based formation coalition model is developed to capture the structural characteristics of formations and the mental and behavioral characteristics of flights under incomplete information. Third, a Bayesian negotiation algorithm is constructed, within which Harsanyi transformation is used to transform the formation coalition problem under incomplete information to a Bayesianequivalent coalition problem under imperfect information. Experiments indicate that the model proposed in this study is fast convergent and produces an equitable formation flight economy split among fleets. Besides, the unknown information prediction accuracy is better than MAS cooperative coalition model. KeywordsCommercial aviation; Formation flight; Coalition; Negotiation; BDI; Incomplete information INTRODUCTION The increasing global air traffic demand in commercial aviation sector not only aggravates the air traffic delay, but also creates more serious energy and environment problems. Research conducted by the International Airport Association indicates that the passenger demand is expected to reach 9.1 billion and cargo demand 214 million tons in 2025, which in turn will result in 1.4 billion tons of CO2 emissions, increasing concerns for energy demand and environment crisis. In 2009, the European Union urged its member states to cut down CO2 emissions to half of the 2005 level by 2050. The aviation sector will inevitably to take strategies to run down its share of CO2 emissions. Flying in formation like migrating birds saving energy, the amount of which varies sharply based on different conditions but is very impressive, over long distances was suggested by many scholars. NASA, Airbus, Boeing and some researchers have pioneered studies on aerodynamic basics and fuel saving of formation flying in commercial aviation community. Commercial formation flight is recommended to be patterned into echelon form with the longitudinal spacing varying from 5 to 40 spans, termed as extended formation flight. Leader shall be rotated among all flights with the optimized interval to make sure each will achieve a fair fuel saving. There are many interesting issues in this area, e.g., aerodynamic coupled effects., precision wake detection and estimation, peak-seeking control based on drag reduction, formation coalition, formation path planning, and etc. However this study only addresses the formation coalition problem. Formation coalition is interpreted as when, where and with who flights are planned to join and break away from a formation, with the objective of maximizing the overall fuel savings. However, formation paths shall be created in advance to evaluate the fuel economy of each specific formation. Therefore, the formation coalition problem and the formation path planning problem are highly correlated and NP-hard. Ribichini formulated the problem as three related subproblems, presented a multi-agent coalition algorithm and solved it via the greedy method. Kent built a mixed integer programming model for large-scale formation coalition and solved it based on simulated annealing. Later, he incorporated wind impacts into the model. Xu developed a bi-level formation flight path planning framework in which heterogeneous aircraft drag models are involved. He also significantly reduced the problem’s complexity by restricting the search space inside the intersections of all the candidate flight performance and fuel-efficiency envelopes. Xu and Meng presented a mathematical model of the formation path planning problem along with related geometric deductions. MENG, Xu and Zhao developed a Multi-agent System (MAS) model addressing the commercial formation coalition problem under incomplete information . This thesis is organized as follows: Section 1 introduces present research achievements of formation flight in commercial aviation community. Section 2 builds the basic MAS framework to fit well agent-based computation. Section 3 implements the BDI-agent model and develops an agentbased negotiation algorithm under incomplete information. Experiments are made in Section 4 to validate the efficiency of the BDI-agent based formation coalition model and negotiation algorithm. Conclusions and suggestions for future work are made in Section 5. MATHEMATICAL FORMULATION A. Problem Formulation In our previous work, the formation flight coalition problem was formulated as an WGSMT construction problem. The formation path can be represented by a WGSMT tree, Ƴ(D,R,B,A,W), spanning the departure aerodrome set, D={di|i=1,2,...,m}, and the arrival aerodrome set, A={aj|j=1,2,...,n} (Fig 1). The rendezvous node set, R={ri|i=1,2,...,m-1}, include a series of optimized nodes after which fleets from different departure nodes or previous rendezvous nodes fly in formation. The breakaway node set, B={bj|i=1,2,...,n-1}, is a series of optimized Steiner nodes, paired with rendezvous nodes, after which fleets leave the formation for their own arrivals or the subsequent breakaway nodes. W is the arc weight set determined by fleet size. The objective is to minimize the total weighted geodesic distance of Ƴ(D,R,B,A,W) by optimizing the formation schedule.