An agent-based infrastructure for monitoring aviation pilot's situation awareness

It has been widely documented that at least 70% of commercial aviation incidents in the last 15 years are to be ascribed to human errors, although that is to be linked together with the difficulties in interactions with the ever more complexity of modern planes. Therefore there has been an enormous increase in avionics and on-board systems which, taken one by one, should increase safety, but on the other hand, often without considerations for the ergonomics of the interfaces used, the whole system leads to an overall increase in the workload and a compromise of the global situation awareness. To tackle this problem, in this paper we present the model of an infrastructure which integrates intelligent agents in order to monitor in real time the attention paid by aviation pilots during training/operative flight missions, to make the decision process easier and increase Situation Awareness (SA). To achieve this goal, from the methodological point of view, in our work we proceed reversing the terms of the problem. In other words, we use the most advanced technology to build an agents-based infrastructure to interpret the reality in which the pilot is set. The model of the infrastructure, we call Situation Awareness Monitoring Infrastructure (SAMI), is based on distributed agents, that cooperating with each other act like a virtual co-pilot in order to augment capabilities of the real pilot, enhancing his SA, and to maintain and recover pro-actively his attention. Starting from an event actually occurred, a case study scenario is given to prove the enhancement given by SAMI in pilot's SA and thus in flight safety.

I. Introduction

Data from the Flight Safety Foundation (FSF) [1] show how the ratio of incidents which have led to the total loss of aircraft (Jets) for commercial transportation in the last 25 years has remained essentially stable, at around 1.5 incidents per million departures. With the doubling of air traffic predicted for 2020 the total number of incidents will rise, even though the ratio of incidents/flight hours will stay low, and this will impact on the passengers' perception of the safety of air transport, as they take more notice of the number of incidents than their ratio. And it has been widely documented that at least 70% of commercial aviation incidents in the last 15 years are to be ascribed to human errors [2], [3], although that is together with the difficulty linked to the interaction with ever more complex planes. This difficulty was confirmed by the industry sector [4] and by later studies [5], which highlight how on the one hand there has been an enormous increase in avionics and on-board systems which, taken one by one, should increase safety (FMS, Narrow spacing for VHF frequency, TCAS); on the other hand, budget requirements, the need to retrofit numerous old-generation aircraft (forced to suffer invasive technological upgrading to be able to use air-spaces efficiently and economically), have produced the installation on board of low cost apparatuses and with often in-existent considerations of the ergonomics of the interface used, leading to an overall increase in the workload and a compromise of the global Situation Awareness. In this paper we will propose an agent-based model for an integrated hardware/software infrastructure able to increase flight safety pro-actively and to anticipate the onset of problems which can lead to incidents. The infrastructure aims also to make easier the decision making process which will lead to a positive solution of the problem. To attain the goal, we will consider the most recent technology in the field of artificial vision and of the measurement of psychophysical parameters, starting from the most recent know ledge of visual attention to arrive at the development of an original and innovative model of augmented reality (for details on augmented reality see [6]). From a methodological point of view, we will propose the development of new closed loop technologies by means of the computerization of the environment in which pilots find themselves operating, so that the system can identify attention lapses, analyze potentially dangerous situations and produce alert signals which allow correct decision making and/or full situation awareness. In this sense the proposed agent-based infrastructure will be integrated with the on-board instruments, and it will be linked with optoelectronic and psychophysiological sensors and other devices that produce useful information about the physical environment in which pilots find themselves operating. The rest of the paper is organized as follows: section II briefly presents the state of the art of both application domain and cognitive approach to the situation awareness; section III shows the agent-based infrastructure model; section IV we will demonstrate the effectiveness of our system model using case study, based on an event actually occurred. Finally section V reports our conclusion and future works.