Robust navigation, which is critical for an animal’s survival, requires the processing of complex
sensory information spanning different modalities and time scales. Unlike human-engineered
systems, where sensors are passive and modularized and decisions are typically made
centrally, biological sensors constantly interact and influence each other, and behavioral
decisions are made on different time scales with diverse goals. Further, such decisions are
based on actively collected sensory information. Our team’s long-term goal is to elucidate the
entire neural circuit dynamics from inter-sensory interaction and multisensory integration in the
central brain to the generation of motor commands across different time scales. We use
Drosophila melanogaster to take advantage of the genetic tools, rich behaviors, and the
available tools to reconstruct the neural circuits with full brain electron microscopy data. We will
investigate how visual information, mechanical information (wind), and gyroscopic information
(sensing body rotation) are integrated and used to generate motor commands. We aim to test
the following hypotheses: (1) Sensory information from one modality affects the processing of
different sensory modalities (e.g., gyroscopic sensation may drive neck muscles to effectively
change the visual input). (2) Information from multiple sensors is integrated in the central brain
using attractor dynamics that unify winner-takes-all and Kalman filter mechanisms. (3) The
central brain not only integrates sensory stimuli, but actively drives sensory muscles to
maximize task-relevant information, increasing robustness of sensory processing. To test these
hypotheses, we combine our team’s diverse and complementary expertise, including two-
photon calcium imaging, one photon muscle imaging, whole-cell patch clamping, precise
sensory perturbation, high speed behavioral analysis, aerodynamics, single-cell resolution
optogenetic perturbation, computational modeling, network theory, and control theory. We will
develop a new control theoretical framework supported by anatomical, behavioral, and
physiological experiments. We will test predictions of models using behavioral and optogenetic
perturbation methods to further refine the theory. The successful completion of this project will
put our team in an ideal position to further investigate multiple sensorimotor transformation
pathways with different time scales, spanning reflexive response (fast), obstacle avoidance
(medium), and voluntary navigational decisions (slow). Overall, our team aims to reveal the
computational principles of neural network dynamics underlying robust navigation.
