Flight Evaluation of an Adaptive Velocity Command System for Unmanned Helicopters

notably the US Department of Defense Predator, have achieved this level of automation and are now being operated on a regular basis. This paper examines the general requirements for fully-autonomous unmanned air vehicle in-flight mission planning and management, sensor and data fusion, synthesis of situation awareness, path planning, guidance, and control in order to define a reasonable set of near term guidance and control system requirements for a U.S. Army unmanned helicopter test-bed. Development of a near term operational capability dictates simplification of the postulated architecture down to the level of “intelligent” way point-following combined with high-level interaction with a human “supervisor.” Practical implementation leads to the development of a velocity command system driven by a relatively simple rule-based command generator. The velocity loop is based on inversion of a very simple helicopter dynamic model at hover, and produces a set of attitude commands. This outer loop is augmented with a neural network to provide adaptation to parameter and environmental uncertainty. A previously developed high-bandwidth adaptive attitude command system is employed for the inner loop. In both the inner and outer loops, pseudo control hedging is used to enable correct adaptation despite periods of control saturation. A coupled analysis of the inner and outer loops is provided for gain selection. The developed velocity command system is implemented in real-time, and evaluated both in high-fidelity nonlinear simulation and in flight test. There is much research being conducted to achieve ever greater levels of “autonomy” in UAV operations, with the ultimate goal being large numbers of fully-autonomous UAV systems performing a variety of complex tasks given only very high-level mission objectives and supervision. Systems under development are both large and small, including representatives such as DARPA’s unmanned combat air vehicle (UCAV) and DARPA’s micro air vehicles. The large systems such as UCAV are being developed to carry sophisticated sensor and weapon systems great distances, and warrant the development of the software required to achieve full autonomy. This task is aided by the fact that these systems are designed for very specific and well-defined missions. However, great need exists for very simple low-cost systems to serve the small unit or even the individual soldier. In this case the applications will be quite varied (and thus are poorly defined at the outset), and sufficient resources will not be available to develop software tailored to each possible scenario. The payload for these systems is most often a simple imaging package for surveillance and reconnaissance tasks. Rotorcraft offer unique flight capabilities that make them especially well suited to this type of work in a very constrained or cluttered environment, and are also receiving a lot of attention. For example the U.S. Army has near-term plans to evaluate the use of unmanned rotorcraft in a variety of applications using several unmanned helicopter testbeds. Operations of interest include tasking and supervision of the unmanned rotorcraft by the crew of manned flight systems. INTRODUCTION For many years now, researchers and engineers have been working toward the design and development of unmanned air vehicles (UAVs) that are sufficiently automated to enable management by a remote human operator who is relieved of actual piloting duties. A number of UAV systems, most In this paper we consider the elements of a generic guidance and control system for an unmanned rotorcraft that can be configured at varying levels of sophistication to support a wide variety of potential tasks. We then identify a specific subset of tasks we would like to perform in the near term, and the associated guidance and control system requirements. The paper then proceeds with development of an adaptive velocity command system to meet these ____________________________ * President, Member AIAA † Professor, Fellow AIAA ‡ Professor, Member AIAA †† Graduate Student ‡‡ Research Engineer American Institute of Aeronautics and Astronautics 1 requirements. The velocity command system is driven by a relatively simple rule-based command generator. The resulting velocity command system is combined with a previously developed adaptive attitude command system, and then evaluated both in nonlinear simulation and flight test. AUTOMOUS UAV ARCHITECTURE Figure 1 presents a generic architecture for an autonomous UAV mission planning, mission management, path planning, guidance, and control system, along with associated elements for synthesizing what is often referred to as “situation awareness.” There are three basic functional groups depicted. The upper portion (shown in green) represents elements designed to fuse information and requirements into a plan of action. The middle blocks (shown in blue) are associated with path planning and guidance. The lower blocks (shown in grey) are for aircraft-specific control functions. Also shown in the figure (in yellow) are some of the vehicle dynamic and sensor models required to simulate operations. The components of this system and their interconnections are purely notional, and were not derived from any current specification or requirement. At the very top is a block that represents a command and/or supervisor interface. This could represent the combination of a preplanned set of mission objectives and a traditional UAV operator. Alternately this block could include a natural language interpreter/translator and an associated intelligent agent to enable direct supervision of the UAV in an interactive manner by personnel whose primary responsibility is other than UAV operation (e.g. a crew member on a manned aircraft, or a soldier on the ground). In the case of a fully-autonomous UAV, operating either alone or cooperating with other autonomous assets in some fashion, this block may represent the interface to other hardware/software systems. The next layer down is a high-level software component for in-flight mission planning and updating, the latter process being referred to as mission management. This block translates all of the inputs into a consistent set of objectives, requirements/constraints that are representative of the pre-mission plan and/or supervisor’s intent. Depending on the specifics of the system, data from external sources and on-board sensors is next fused and integrated into a current world model, which is combined with the interpreted mission objectives and constraints to produce some degree of “situation awareness.” The output of this process is a current plan for motion of the vehicle/sensor(s), including consideration of dynamic constraints. The motion requirements are then passed to one or more guidance laws that employ the attitude command system to produce the commanded motion.