In Chapter 5, I describe the role of the neuromodulator octopamine in the flight control system, and use a control theoretic framework to propose a putative neural circuit for the flight speed regulator. Perhaps the most elegant example of stigmergy can be found in the nest-building behavior of the solitary mud wasp [132].
Abstract
These results provide a quantitative behavioral background for elucidating the neural basis of plume tracking using genetic and physiological approaches.
Introduction
Another critical unknown is the extent to which flies change their behavior in the face of different wind speeds, an environmental condition that varies widely in the field. These experiments provide a comprehensive and detailed overview of the algorithm that flies use to locate an odor source, and can be described in the context of a simple stigmergic model.
Methods
- Animals
- Flight arena
- Odor stimuli
- Odor delivery
- Odor plume calibration
- Trajectory reconstruction and analysis
- Model
- Statistics – Fisher’s exact test
- Statistics - bootstrapping
Under this assumption, the mole fraction of ethanol in air is described by the ratio between the vapor pressure of ethanol and the atmospheric pressure:. These p-values give a quantitative statistical sense of the significance of the behavioral differences between the clean air and attractive odor cases.
Results
Surging behavior
In the presence of an attractive wind, flies showed a clear preference for flying upwind compared to downwind (Fig. 2.5B), consistent with previous studies. Flies also showed a significantly narrower distribution in the presence of faster wind speeds (p≤0.001, Fischer's exact test) and a reduced accuracy in moving upwind at slower wind speeds (Fig. 2.6 C), results that are also consistent with visual anemotaxis.
Casting behavior
This hypothesis was supported by my experimental results, which show a significantly (p≤0.001, Fisher's exact test) tighter crosswind direction distribution when the lines are perpendicular to the wind (Figure 2.8B). The difference in the histograms shown here and in Fig. 2.8A is expected, since the auxiliary body orientation camera was centered and could only cover a small part of the width of the wind tunnel. ii) Title distribution vs.
Odor induced visual saliency
-B) Height response of flies relative to the time they leave the odor plume for different visual environments, for the duration before re-entering the plume. The light red cover in the ethanol container indicates the extent of the odor plume.
Discussion
- Casting, surging, and odor induced object salience constitutes
- The role of odor induced visual saliency
- Plume tracking, and visual saliency, do not diminish over time 39
- Conclusion
The relative timing of the vertical and horizontal aspects of shedding has important implications for the overall research pattern that flies execute in the upwind plane. The tracking behavior of flies can be described by a simple stigmergic model consisting of three distinct reflexes: (1) flies rise upwind within 270 ms of entering a wind plume: (2) upwind released 640 ms after loss of the plume and (3) explore high-contrast visual features in the vicinity of the plume.
Supplementary figures
Each panel shows a heatmap representation (log color scale) of the per-pixel p-values describing the likelihood that the behavioral differences I observed in the presence of ethanol versus clean air are due to random sampling. Thrown course in crosswind plane, see figure E) Height response with checkerboard floor, see figure F) Height response with low contrast floor and dots, see figure. This analysis was made possible by a custom feedback system I built that actively kept the fly in the focus of the high-speed camera.
Finally, the fly extends its legs when the visual target reaches a threshold retinal size of about 60°.
Introduction
Together, the results provide insight into the organization of sensory motor modules underlying insect landing and foraging behaviors. Although the landing sequences of Drosophila free flight have not been explicitly studied before, several behavioral modules that likely represent behavioral components have been extensively studied. If these saccades are directed toward certain features, they should also be considered an early component of the landing sequence.
Although this reflex has previously been termed the “landing response,” it represents only one component of landing behavior and its position in the complete free-flight landing sequence is unknown.
Methods
- Animals
- Flight arena
- Experiment protocol
- Trajectory reconstruction and analysis
- High speed imaging
- Analysis of saccades
- Procedures for analyzing landing behavior
- Statistical analysis
The post angle (φ) is defined as the azimuthal angle of the center of the post from the flight path. Fly-bys were analyzed only from the start of the trajectory to the point just before the first saccade following the closest approach to the pole (see Figure 3.2). The camera was activated afterwards whenever a fly came within 1 cm of the pole (both landings and close fly-bys were recorded).
The distributions show the ratio of the sum of �Hs to the sum of �Hs+�Hn for each trajectory.
Results
- Description of landings and flybys
- Saccade results
- Landing behavior
- Crash landings
- Post texture
This suggests that saccades made by flies in the vicinity of the post are likely to be responses to the post, especially when the retinal size exceeds 25°. After turning, flies show a qualitatively different post angle distribution in the presence (black) and absence (green) of the post. Flies show a qualitatively different retinal size distribution at which they make their last saccade in the presence (black) and absence (green) of the drug.
To further investigate the aversive saccades, I again set a retinal pole size threshold at 25° (corresponding to a distance of approximately 3 cm).
Discussion
- Attractive and aversive saccades
- Landings
- To land, or not to land?
- Crashes
- Post texture
- Summary
Given the limitations of the visual and mechanosensory systems available to the fly, it is more likely that they use a measure of the retinal size of the pole (which is correlated with distance) and the rate of expansion (which is correlated with ground speed). The time between leg extension and landing is less than 50 ms for about 1/3 of the landings I have observed (Fig. 3.15). In most touchdowns I noticed that one of the two front legs would touch the post before the other.
Although subtle, the differences in behavior in the presence of the checkered and plain black posts merit comment.
Supplemental figures
To increase the number of trajectories in these control data sets, I extended the imaginary post to the top of the arena. Saccade turn angle (ψ) is plotted as a function of follow angle (φ) for landings (A) and flybys (B), with the color of each point indicating the retinal size of the pole at the time of the saccade. In panels C and D, the associated shading shows smooth representations of the distributions calculated with 3rd order 0.3 Hz Butterworth filters.
Although the statistics suggest that there may be significant differences in deceleration behavior due to post texture for the flies that did not turn after initiation of deceleration (purple dots, p(x2)=0.66, p(x1:x2)=0.011) , visual inspection of the data shows no appreciable difference.
Introduction
Review of visual distance estimation in biological systems
Stereopsis works by calculating the parallax between two (or more) simultaneous images of the same object from different viewpoints to triangulate the absolute distance to the object. Then look at the visual system of the fruit fly, which has an interocular distance of about 0.3 mm. Although the exact angular resolution of the flies' visual system is not known, I approximate it by the ommatidal acceptance angle of 5°, which yields the red curve in Figure 4.1.
This approach, which I call dynamic observation, has the intuitive functionality of a nonlinear observer, which I will develop more formally in the following sections.
Modeling and observability analysis
Problem statement
Analysis
To use linear system analyses, I start by linearizing the system over a nominal trajectory, (dt(t), vt(t)). To address the question of whether it is possible to estimate distance and velocity using only optical flow, I investigate the system's observability (a measure of how well the states of a system, such as position and velocity, can be inferred given the available sensory measurements [112]). Instead, it is possible to numerically estimate the observability Gramian, called the empirical local observability Gramian, by simulating the system and comparing the outputs y for perturbations ±� of the initial state [75, 88].
If the Jacobian of O has full rank (that is, if the number of linearly independent terms is equal to the number of states in the system) in all states, the system is said to be observable.
Implementation
Optical flow as a function of camera pixels from two consecutive images, calculated using OpenCV's Lucas Kanade algorithm. For control purposes, I calculated a linear fit of the data (red line) over the region of interest indicated in (A). To use these optical flow estimates for my estimation problem, I first had to calibrate the system, as the Lucas Kanade algorithm provides normalized values between ±1.
I then implemented a simple proportional controller 4.16 with gain k = 6 to adjust the acceleration of the camera so that it maintained the desired value of the optical flow.
Results and discussion
Applications to robotic systems
For example, a flying robot could periodically approach the ground below it with constant optic flow to estimate its altitude and use that measurement as a calibration for other optic flow estimates. Recall that optical flow itself can provide relative measurements for different objects, so that if the distance to one of these objects is known, the others can also be calculated. The optical current can be calculated with simple parallel analog circuits without the need for extensive memory [136].
For simplicity, I have limited my analysis to a single-degree-of-freedom system, but the principles presented here can be generalized to three-dimensional motion.
Implications for landing insects
However, flies that approached the target but did not brake did not extend their legs. These results from tethered flies seem contradictory to my hypothesis because the flies stretch their legs without slowing down physically. But in these tethered flight experiments, it is impossible to know whether the flies were trying to slow down.
Our analysis focused on a single degree of freedom trajectory, which is consistent with the landing behavior of fruit flies that do not make significant changes in heading after the pre-landing delay begins [ 150 ].
Summary
These results suggest that leg extension is not triggered solely by object size or frame rate, but rather depends on some aspect of the flies' internal state or their deceleration behavior, an observation consistent with my model for distance estimation. Tethered flight experiments in fruit flies and other insects have shown that visual stimuli are sufficient to elicit strong leg extension responses [ 17 , 144 ]. It is also possible that there are multiple sensory-motor pathways that can cause leg extension.
Recent evidence suggests that the sensitivity of flies to large-field optic flow is enhanced by the release of octopamine during flight.
Introduction
Local optic flow is estimated by a two-dimensional array of so-called elementary motion detectors and then integrated into visual space by large interneurons in the lobule and lobule plate. Lobule plate tangential cells (LPTCs) are particularly well characterized, and many display receptive fields that make them sensitive to various patterns of self-motion, such as those generated by rotation and translation in flight. Additional experiments in which lobule plate neurons are removed by physical or genetic means add further support to this hypothesis.
In the case of flight, this modulation appears to be mediated by octopamine neurons which become active during flight and cause an increase in the gain of visual responses in at least one class of LPTCs, vertical system (VS) cells [142].
Methods
Animals
Flight arena
Experiment protocol
Trajectory reconstruction and analysis
Immunohistochemistry and imaging
Across all genetic lines and all time frequencies, over 92% of the flies' speed was in the direction of visual movement for the duration of the trials. A motion-selectively detecting neuron in the locust brain: Physiological and morphological characterization. An investigation of the mechanisms underlying nest construction in paralastor sp.
Effect of visual landscape on free-flight behavior of the fruit fly Drosophila melanogaster.
Wind tunnel and odor plume calibration
Flight trajectories in the presence of an attractive odor plume
Characterization of flies’ responses to an ethanol plume
Time flies spend inside an odor plume, and between plume encounters. 23
Surge behavior is a function of both visual and wind stimuli
Flies cast crosswind within 600-800 ms after leaving an attractive odor
Casting behavior is a function of visual, but not wind, stimuli
Flies cast with a range of slip angles
Slip angles during casting are a function of flight speed
Flies are attracted to high contrast visual features
Flies are attracted to high contrast visual features (II)
A simple model of odor tracking behavior suggests that surge, cast, and
Summary diagram, indicating the three independent sensory motor mod-
Behavioral responses to a plume of ethanol and Vector960 (fruit fly
Behavioral responses to a plume of ethanol and Vector960 (fruit fly