MAVs and Fruit Flies: Unguided Evolution Smarter Than Top Scientists

Autonomous air vehicles are finding increasing use and the Air Force is interested in micro versions:

Micro Air Vehicles (MAVs) typically UAVs with wingspan on the order of 15cm or less are fast becoming commonplace for meeting a wide range of current and future military missions.

But there are tremendous technical challenges:

A typical sensor suite for a MAV consists of GPS, MEMs-based linear accelerometers, angular rate sensors, magnetometers, and barometric-altimeters. While this is adequate for waypoint navigation, the potential of MAVs to replicate the flight agility of natural fliers (e.g., birds, bats, insects) remains elusive, especially in complex terrain such as city streets or forests.

For hints at how to solve such problems designers are looking at nature’s solutions:

The desire to engineer the agility of natural fliers has led researchers to the study of flying organisms to learn how animals combine sensory input with control output to achieve flight maneuverability. Biologists are beginning to understand how visual information is integrated with mechanosensory information in biological systems for flight stabilization, landing, and prey/mate pursuit. Studies are also underway to discover how proprioceptive sensory feedback is used for fine-scale control the movement of wings, legs, etc. during aggressive maneuvers (e.g., obstacle or collision avoidance). These sensory modalities are combined with olfactory or auditory information for predator avoidance and prey/mate pursuit.

Fortunately evolution has created highly advanced flight systems:

The fact that animals such as fruit flies exhibit such remarkable flight agility with many sensory inputs and modest onboard processing suggests a particular kind of coupling between sensing, control and dynamics altogether qualitatively different from that of engineered systems. Advancements in flow control have made it possible to control the separation of flow around wings, either to inhibit separation for higher cruise lift-to-drag ratios or to promote it for large transients in aerodynamics loading for aggressive maneuvers. Natural flyers have anatomic features which probably act as flow control devices (e.g., covert flaps) and may act as aerodynamic sensors.

But understanding evolution’s marvels remains a research challenge:

Rigorous system modeling that can accurately capture the vehicle dynamics, sufficiently accounting for uncertainties in aerodynamic and structural models, remains primitive even for engineered vehicles, let alone for natural flyers. Uncertainty arises both in the veracity of particular models in describing a given flow or dynamics phenomenon, and in unknowns in the inputs, such as wind gusts and their time-dependent effect on the vehicle. While on-going research efforts are addressing some of the critical limitations in this area, significant uncertainties in the dynamics models of MAVs are unlikely to be completely eliminated.

How do random mutations produce such brilliant designs? Answering such questions is, of course, what science is all about. As Darwin explained, evolution opens up wide areas of scientific research. But now we know it also gives top scientists hints to their toughest problems.