Analysis And Design Of TCAS Logic Based On MDP

The Traffic Alert and Collision Avoidance System(TCAS), currently mandated on large transport aircraft, is mainly used for reducing the risk of collision with other aircrafts. TCAS is to guide the pilot to make the maneuver to resolve the risk of mid-air collision between aircrafts. Logic unit is an important module of TCAS systems. According to the relative range and relative height of an intruder and the own aircraft, logic units determine whether the intruder would be a threat to the own aircraft and then decide whether to issue resolution advice.The logic unit of current TCAS is mainly implemented by pseudo-codes and includes many heuristic rules and couple parameters. To adapt the evolution of airspace and the development of surveillance technology, the logic unit needs to be modified. Due to the complexity of logic unit, modifying the logic unit is very hard and may introduce some new vulnerabilities. This paper presents a new method based on the Markov decision process and it involves using computers to directly optimize the logic unit based on encounter models of traffic and performance metrics. When the airspace model or surveillance system is changed, it only needs to modify the corresponding model and then optimize the logic unit. Therefore, this method can satisfy the requirement of the rapid development of airspace and surveillance technology. The contents of this paper are given as follows.Firstly, the feasibility of developing the logic unit with the Markov decision process is demonstrated and the advantages of this method are presented. The Markov decision processes can not only take account of the uncertainty of the encounter process, but also includes the future states and decision information. The logic unit designed by this method will be more reliable.Secondly, the logic unit is designed and implemented for two aircrafts encounter situation in the two-dimensional space. Cost function is firstly defined with some performance metrics, then the encounter process of aircrafts with MDP is modeled, the optimal policy lastly is determined with dynamic programming and evaluated by logic visualization, simulative encounter and Monte Carlo method.Thirdly, two kinds of probabilistic pilot respond models, i.e. linear pilot respond model and quadratic pilot respond model, are considered. The linear pilot respond model is that pilot can respond to the advisory immediately or choose to ignore all advisories. The quadratic pilot respond model is that the pilot can respond to the advisory immediately or continue following the current advisory. The two proposed logics are evaluated.Finally, a model is built for both aircrafts equipped with collision avoidance systems and coordinated by the two advisors. Coordination generally can greatly reduce the risk of collision. The logic unit is evaluated for the synchronous action set and the deterministic pilot response model.