Evaluation of Possible Flight Strategies
363
4.2 Technical Implementation
This section shall only discuss possible implementation methods. Nothing discussed in
this work has been implemented yet. Cost functions are frequently used in technical
applications to choose between several possibilities, also in some flight controllers. The
obstacle independent linear regression curve of the format:
HV = 0.056 d − 1.2732
(2)
can serve as an element of a flight controller, in which the HV index can be understood
as a bias or tendency to a specific avoidance direction of the flight control system.
Equation (2) can either be used to define a corridor of possible evaluation points or to
introduce a cost function penalty. In the first case, with increasing positive HV index,
waypoints with decreasing horizontal distance from the current flight path are excluded
from consideration; a negative HV index might lead to the exclusion of waypoints at
specific vertical distances. The advantage of this approach is that the number of evaluated
points is reduced before evaluation, which reduces the computational burden. In the
second case, the HV index regression could serve as an additional penalty in a cost
function. For example, the value a positive HV index can, obtained by Eq. (2), can be
used as an additional penalty for horizontal deviation, while the absolute value of a
negative HV index can be used as a penalty for vertical deviation. However, the specific
values of the regression line need to be adjusted to the aerodynamic characteristics of
the flying platform.
While the HV index might improve overall performance, it also bears two disadvantages. First of all, any HV index regression control is only as reliable as the distance
measurement of the drone. Unfortunately, the accuracy of the distance measurement
system often scales with its weight; high accuracy is often only available at the cost of
relatively high weight. Secondly, the HV index regression cannot be used as some kind
of gain. Otherwise, the evasion strategy would change with decreasing distance to the
obstacle. Additional methods are required to ensure that the evasion strategy is fulfilled
as intended. One possibility is to plan several waypoints ahead and only change the
planning for safety-critical reasons.
5 Conclusion
This work analyses the behavior of bumblebees when evading obstacles in different
distances and with different dimensions. It is observed that bumblebees, independent of
the obstacle dimensions or their speed, tend to evade very close obstacles horizontally,
e.g., fly around the obstacle, while they prefer to evade vertically, e.g., fly over the
obstacle, if the obstacle is in greater distance. We found indications that the dimension
of the obstacle might influence the decision of the bee as well. However, further research
is required to give a qualitative statement.
Additional research on the specific reasons for the shown behavior is required to
adequately define a function, which can be transferred in UAV control algorithms. However, one approach to transfer these methods to a technical application was presented. In
this approach, the HV index regression curve can serve as a measure to exclude specific
regions of the field of view for consideration of a cost function.
363
4.2 Technical Implementation
This section shall only discuss possible implementation methods. Nothing discussed in
this work has been implemented yet. Cost functions are frequently used in technical
applications to choose between several possibilities, also in some flight controllers. The
obstacle independent linear regression curve of the format:
HV = 0.056 d − 1.2732
(2)
can serve as an element of a flight controller, in which the HV index can be understood
as a bias or tendency to a specific avoidance direction of the flight control system.
Equation (2) can either be used to define a corridor of possible evaluation points or to
introduce a cost function penalty. In the first case, with increasing positive HV index,
waypoints with decreasing horizontal distance from the current flight path are excluded
from consideration; a negative HV index might lead to the exclusion of waypoints at
specific vertical distances. The advantage of this approach is that the number of evaluated
points is reduced before evaluation, which reduces the computational burden. In the
second case, the HV index regression could serve as an additional penalty in a cost
function. For example, the value a positive HV index can, obtained by Eq. (2), can be
used as an additional penalty for horizontal deviation, while the absolute value of a
negative HV index can be used as a penalty for vertical deviation. However, the specific
values of the regression line need to be adjusted to the aerodynamic characteristics of
the flying platform.
While the HV index might improve overall performance, it also bears two disadvantages. First of all, any HV index regression control is only as reliable as the distance
measurement of the drone. Unfortunately, the accuracy of the distance measurement
system often scales with its weight; high accuracy is often only available at the cost of
relatively high weight. Secondly, the HV index regression cannot be used as some kind
of gain. Otherwise, the evasion strategy would change with decreasing distance to the
obstacle. Additional methods are required to ensure that the evasion strategy is fulfilled
as intended. One possibility is to plan several waypoints ahead and only change the
planning for safety-critical reasons.
5 Conclusion
This work analyses the behavior of bumblebees when evading obstacles in different
distances and with different dimensions. It is observed that bumblebees, independent of
the obstacle dimensions or their speed, tend to evade very close obstacles horizontally,
e.g., fly around the obstacle, while they prefer to evade vertically, e.g., fly over the
obstacle, if the obstacle is in greater distance. We found indications that the dimension
of the obstacle might influence the decision of the bee as well. However, further research
is required to give a qualitative statement.
Additional research on the specific reasons for the shown behavior is required to
adequately define a function, which can be transferred in UAV control algorithms. However, one approach to transfer these methods to a technical application was presented. In
this approach, the HV index regression curve can serve as a measure to exclude specific
regions of the field of view for consideration of a cost function.