Evaluation of Possible Flight Strategies
355
from drug delivery to inspection tasks. Stressful times with social distancing, like the
pandemic spread of COVID-19, boost the necessity of autonomously flying delivery
and surveillance drones. However, several challenges remain until an autonomously
operating UAV service is established. One of these challenges is the development of a
reliable obstacle avoidance system. Even though the field of bio-inspired flight control
systems, especially the applied sensor technology, advanced recently [1].
We present possibilities and approaches to improve current obstacle avoidance algorithms, which are found by the analysis of obstacle avoidance strategies used in nature.
More particularly, the flight behavior of bumblebees, Bombus Terrestris, is investigated,
when encountering differently sized obstacles in different distances. The analysis aims
at the derivation of fundamental strategies for improved obstacle avoidance. It is of particular interest, in which situations a vertical evasion is more desired than a horizontal
evasion. The next section will present the theoretical background of this work, as well
as the experimental set-up, which was used to extend the current body of knowledge.
Section 3 will give an overview of the results from the conducted experiments, while
Sect. 4 contains the conclusions drawn from the results. Finally, the last chapter will
summarize the conclusions.
2 Background
Neuroethologists have studied the behavior of animals for decades. Their research
revealed a variety of strategies on how animals navigate in different environments [2],
avoid obstacles [3], optimize paths [4], and solve other problems [5, 6]. The study of
insect behavior yield applications in technical systems such as network systems [7]
and autonomous robots [8, 9]. Insects and, more precisely, bumblebees are astonishing
foragers, which have been subject of investigation for decades.
2.1 Biological Findings and Their Possible Technical Application
Honeybees and bumblebees forage between rewarding food sources and their hive in
often complex environments. They have to fly through clutter consisting of obstacles of
different sizes, shapes, orientations, and textures. To avoid physical damage [10] and to
perform the task of food collection as efficient as possible, insects need to go around
objects obstructing their way [11]. Flying insects, in particular, can apply a variety
of strategies and flight maneuvers to avoid objects and reach their goal efficient, fast,
and safe [12–14]. Flying technical systems (“drones”) have the same requirements and
encounter the same challenges. When following a global path to a specific goal position,
a flying platform will encounter different obstacles [15]. After obstacle detection, a local
path, evading the obstacle, has to be determined. While several systems and methods
for obstacle detection are known, e.g., LiDAR, radar, or camera with SLAM or SURF
[16, 17], only a few strategies exist for the 3D obstacle avoidance maneuver. The p×4
avoid algorithm, which discretizes the world and evaluates different possibilities with a
cost function, is one of the most sophisticated algorithms for obstacle avoidance [18].
However, this algorithm relies on one core function for all situations and does not adapt
to different situations [19]. Another problem of optimization methods is the definition
355
from drug delivery to inspection tasks. Stressful times with social distancing, like the
pandemic spread of COVID-19, boost the necessity of autonomously flying delivery
and surveillance drones. However, several challenges remain until an autonomously
operating UAV service is established. One of these challenges is the development of a
reliable obstacle avoidance system. Even though the field of bio-inspired flight control
systems, especially the applied sensor technology, advanced recently [1].
We present possibilities and approaches to improve current obstacle avoidance algorithms, which are found by the analysis of obstacle avoidance strategies used in nature.
More particularly, the flight behavior of bumblebees, Bombus Terrestris, is investigated,
when encountering differently sized obstacles in different distances. The analysis aims
at the derivation of fundamental strategies for improved obstacle avoidance. It is of particular interest, in which situations a vertical evasion is more desired than a horizontal
evasion. The next section will present the theoretical background of this work, as well
as the experimental set-up, which was used to extend the current body of knowledge.
Section 3 will give an overview of the results from the conducted experiments, while
Sect. 4 contains the conclusions drawn from the results. Finally, the last chapter will
summarize the conclusions.
2 Background
Neuroethologists have studied the behavior of animals for decades. Their research
revealed a variety of strategies on how animals navigate in different environments [2],
avoid obstacles [3], optimize paths [4], and solve other problems [5, 6]. The study of
insect behavior yield applications in technical systems such as network systems [7]
and autonomous robots [8, 9]. Insects and, more precisely, bumblebees are astonishing
foragers, which have been subject of investigation for decades.
2.1 Biological Findings and Their Possible Technical Application
Honeybees and bumblebees forage between rewarding food sources and their hive in
often complex environments. They have to fly through clutter consisting of obstacles of
different sizes, shapes, orientations, and textures. To avoid physical damage [10] and to
perform the task of food collection as efficient as possible, insects need to go around
objects obstructing their way [11]. Flying insects, in particular, can apply a variety
of strategies and flight maneuvers to avoid objects and reach their goal efficient, fast,
and safe [12–14]. Flying technical systems (“drones”) have the same requirements and
encounter the same challenges. When following a global path to a specific goal position,
a flying platform will encounter different obstacles [15]. After obstacle detection, a local
path, evading the obstacle, has to be determined. While several systems and methods
for obstacle detection are known, e.g., LiDAR, radar, or camera with SLAM or SURF
[16, 17], only a few strategies exist for the 3D obstacle avoidance maneuver. The p×4
avoid algorithm, which discretizes the world and evaluates different possibilities with a
cost function, is one of the most sophisticated algorithms for obstacle avoidance [18].
However, this algorithm relies on one core function for all situations and does not adapt
to different situations [19]. Another problem of optimization methods is the definition