Evaluation of Possible Flight Strategies for Close
Object Evasion from Bumblebee Experiments
Andreas Thoma 1,2(B) , Alex Fisher 2 , Olivier Bertrand 3 , and Carsten Braun 1
1 FH Aachen, Hohenstaufenallee 6, 52064 Aachen, Germany
a.thoma@fh-aachen.de
2 RMIT University, Bundoora, VIC 3083, Australia
3 Universität Bielefeld, 33615 Bielefeld, Germany
Abstract. The need for robust and efficient obstacle avoidance algorithms for flying platforms increases at a fast pace. Strategies are required to maneuver around
various obstacles. Since bumblebees are efficient fliers that need to navigate in
cluttered environments, they are perfectly suitable test objects when seeking for
avoidance strategies. In the present work, we study the maneuver of bumblebees
confronted with a large rectangular obstacle lying on their direct path to the hive.
The bumblebees could always evade the obstacle vertically or horizontally, e.g.,
fly over the obstacle or fly around the obstacle, respectively. The chosen evasion
maneuver, i.e., the basic strategy employed by the bee, is considered the dependent variable. The influence of the distance to the obstacle, obstacle dimensions,
acceleration, and flight speed were investigated and considered as independent
variables. To evaluate the bumblebee behavior, linear regression, and the Horizontality Verticality Index (HV), an adaption of the Laterality Index, was used to
estimate the preferred behavioral choice.
Examination of the bumblebee behavior revealed a strong tendency towards
vertical evasion at higher distances to the obstacle, while the bumblebees evaded
close obstacles horizontally. This is reasonable because climbing in flapping-wing
flight is aerodynamically more efficient in forward movement than in hover flight.
A linear function based on the HV Index was defined to estimate a relationship
between distance to obstacle and evasion maneuver. Depending on the dimensions
of the obstacle, alternative slopes of the HV function could be identified, indicating an additional dependency on the height-to-width ratio. However, taking the
obstacle shape into account does not improve the predictability of an L1 LASSO
regression model.
Finally, one possibility to include the HV index into a technical system as an
element of an obstacle avoidance algorithm is discussed.
Keywords: Obstacle avoidance · Bumblebees · MAV · UAV · Flight control
1 Introduction
Autonomous vehicles are an overall increasing sector. Specifically, possible applications
of unmanned aerial vehicles (UAV) are growing fast. The variety of applications ranges
© Springer Nature Switzerland AG 2020
V. Vouloutsi et al. (Eds.): Living Machines 2020, LNAI 12413, pp. 354–365, 2020.
https://doi.org/10.1007/978-3-030-64313-3_34
Object Evasion from Bumblebee Experiments
Andreas Thoma 1,2(B) , Alex Fisher 2 , Olivier Bertrand 3 , and Carsten Braun 1
1 FH Aachen, Hohenstaufenallee 6, 52064 Aachen, Germany
a.thoma@fh-aachen.de
2 RMIT University, Bundoora, VIC 3083, Australia
3 Universität Bielefeld, 33615 Bielefeld, Germany
Abstract. The need for robust and efficient obstacle avoidance algorithms for flying platforms increases at a fast pace. Strategies are required to maneuver around
various obstacles. Since bumblebees are efficient fliers that need to navigate in
cluttered environments, they are perfectly suitable test objects when seeking for
avoidance strategies. In the present work, we study the maneuver of bumblebees
confronted with a large rectangular obstacle lying on their direct path to the hive.
The bumblebees could always evade the obstacle vertically or horizontally, e.g.,
fly over the obstacle or fly around the obstacle, respectively. The chosen evasion
maneuver, i.e., the basic strategy employed by the bee, is considered the dependent variable. The influence of the distance to the obstacle, obstacle dimensions,
acceleration, and flight speed were investigated and considered as independent
variables. To evaluate the bumblebee behavior, linear regression, and the Horizontality Verticality Index (HV), an adaption of the Laterality Index, was used to
estimate the preferred behavioral choice.
Examination of the bumblebee behavior revealed a strong tendency towards
vertical evasion at higher distances to the obstacle, while the bumblebees evaded
close obstacles horizontally. This is reasonable because climbing in flapping-wing
flight is aerodynamically more efficient in forward movement than in hover flight.
A linear function based on the HV Index was defined to estimate a relationship
between distance to obstacle and evasion maneuver. Depending on the dimensions
of the obstacle, alternative slopes of the HV function could be identified, indicating an additional dependency on the height-to-width ratio. However, taking the
obstacle shape into account does not improve the predictability of an L1 LASSO
regression model.
Finally, one possibility to include the HV index into a technical system as an
element of an obstacle avoidance algorithm is discussed.
Keywords: Obstacle avoidance · Bumblebees · MAV · UAV · Flight control
1 Introduction
Autonomous vehicles are an overall increasing sector. Specifically, possible applications
of unmanned aerial vehicles (UAV) are growing fast. The variety of applications ranges
© Springer Nature Switzerland AG 2020
V. Vouloutsi et al. (Eds.): Living Machines 2020, LNAI 12413, pp. 354–365, 2020.
https://doi.org/10.1007/978-3-030-64313-3_34