A HUMAN FACTORS TESTBED FOR COMMAND AND CONTROL OF UNMANNED AIR VEHICLES

摘要

As technological advances allow unmanned air vehicles (UAVs) to operate more autonomously, human performance in supervisory control of UAVs becomes a critical issue. In particular, experience shows that human performance in supervisory control is a function of extent of automation, levels of system fidelity, and rates of information update. In order to explore this issue, we developed a testbed that supports the planning and conducting of UAV human factors experiments. The testbed is built upon the Multi-Modal Immersive Intelligent Interface for Remote Operation (MIIIRO) which was designed to support UAV control [1]. The testbed implements a client/server architecture in which UAV operations are simulated on a server that maintains the states of the UAVs and their payloads. The operator control interface is implemented as a client Java applet which can be run via a web browser or as a standalone application. This configuration supports the collaboration of a crew of operators in control of multiple UAVs. The clients communicate with the server via message passing with messages that are time stamped and placed in a message queue so that their delivery can be delayed arbitrarily to simulate various link delays. An assortment of facilities have been developed in the testbed client to support both the planning and execution of human factors experiments. An experiment designed by the Air Force Research Laboratory (AFRL) Crew System Interface Division concerns a multiple UAV mission to locate and identify ground targets. This experiment was designed to investigate several human factors issues raised in earlier work [2], including operator interaction with automation levels and decision-aid fidelity. The testbed client enables the experiment planner to design routes for the UAVs by designating waypoints in the Tactical Situation Display (TSD). The client also automates the generation of images of ground targets based on the specified parameters, such as the number of targets in each image, type of targets, and fidelity level of the automatic target recognition (ATR). Another planning facility supported by the client is event generation. It allows the planner to add events such as pop-up threats, ad-hoc targets, unidentified aircraft, and mission mode indicators during the experiments. During an experimental trial, as the UAV flies over a target area, the captured images are entered in the Image Queue. The image at the top of the queue is displayed with boxes surrounding the ATR identified targets. The experiment subject must then verify the ATR and redesignate the boxes (if necessary) optionally within a specified time limit. Events also occur at random times and the subject is required to respond to them. The subject's responses and their timings are recorded bv the client for later analvsis.