Page 14 - Neural Network Modeling and Identification of Dynamical Systems

2 NEURAL NETWORK MODELING AND IDENTIFICATION OF DYNAMICAL SYSTEMS

• to be able to assess the current situation on task is accomplished by some team of aircraft,
the basis of a multilateral perception of the ex- which includes the given UAV.
ternal and internal environment, to be able to The control algorithms (formation of control
form a forecast of the development of the sit- actions, decision making for control) should use
uation;
• to gain new knowledge, accumulate experi- information about the mission goals and about
the situation characterized by assessments of the
ence in solving various tasks, learn from this current and predicted situation in which it per-
experience, and modify its behavior based on
forms the task of the UAV, as input data. This
the knowledge gained and accumulated expe-
situation is made up of both external compo-
rience;
nents (state of the environment, the state and
• to be able to learn how to solve problems not
actions of partners and opponents) and internal
provided for by the original design of the sys- components (data on aircraft state, diagnostics
tem; data, and performance evaluations of the struc-
• to form teams that are able to solve some
ture and aircraft systems). Means of obtaining
problem by interactions between their mem-
this basic information should also be included
bers.
in the complex of algorithms that implement the
In order for robotic UAVs to be able to accom- desired behavior of a robotic UAV.
plish difficult missions on the same efficiency The aforementioned requirements can only
level as a manned aircraft, a radical revision of be fulfilled if the UAV’s behavior control sys-
the current approach to development and man- tem possesses advanced mechanisms, which al-
agement of control algorithms for UAV behavior low an adaptation to significantly changing situ-
is needed. In robotics, the totality of all types of ations with a high degree of uncertainty and also
processes of functioning of the robot is usually learning and knowledge acquisition based on
called the behavior of the robot. Accordingly, current UAV activity for future use. Such mech-
bearing in mind the ever-increasing trends in the anisms should allow the possibility to solve the
robotization of the UAV, it is accepted to talk following important tasks:
about the task of the behavior control for UAVs • obtaining situation awareness which involves
as the implementation of all types of its func- current situation assessment and future situa-
tioning necessary to fulfill the abovementioned tion prediction;
target tasks. The behavior control of the UAV in- • synthesis and implementation of UAV behav-
cludes the following elements: ior as an aggregation of purposeful reactions
to a current and/or predicted situation.
• planning flight operations, managing its im-
plementation, updating the plan when a situ- The implementation of these mechanisms
ation changes; provides the ability to create adaptive and intel-
• UAV motion control, including its trajectory ligent systems to control the behavior of UAVs.
motion (including guidance and navigation) The use of such systems gives an opportunity
and angular motion; to implement highly autonomous robotic UAVs,
• management of the solution of target tasks designed to effectively accomplish difficult mis-
(control of the action of observation and re- sions under uncertainty conditions. Another im-
connaissance equipment, control of the ac- portant implication of adaptive and intelligent
tions for performing assembly operations, control of UAV behavior is the possibility to sig-
etc.); nificantly increase survivability of an aircraft in
• management of interaction with other air- case of severe airframe damage and onboard
craft, both unmanned and manned, when the systems failures.