Blake Barber and Dr. Tim McLain, Mechanical Engineering
A key enabling technology that will allow the U.S. Military to maintain information superiority in the 21st century is the development and implementation of Unmanned Air Vehicle (UAV) technology. Especially critical to the support of the small special operations units conducting the asymmetric warfare of the 21st century will be the development of mini and micro UAV technology. Mini and micro UAV technology will enable foot soldiers to hand launch and recover small, robust reconnaissance aircraft capable of providing real-time video surveillance and real-time target acquisition. These capabilities will allow soldiers to have the full situational awareness they need for maximum effectiveness. Mini and micro UAVs may also be equipped with specialized sensors for early detection of biological and chemical agents. Small swarming UAVs may be used to debilitate enemy equipment by jamming radar and communications or by damaging vulnerable equipment such as jet engines. Mini and micro UAV technology will also extend into nonmilitary applications such as border patrol, search and rescue, hazardous material inspection, crowd monitoring and control, traffic monitoring, and homeland security.
The usefulness of mini and micro unmanned air vehicles depends largely on the ability to operate the UAV in uncontrolled environments. This includes the ability to track terrain and land in environments where the terrain is not fully known. The development of mini and micro UAVs with the ability to accomplish these tasks has been most hindered the lack of suitably small and reliable sensors for detection of height above ground. Our approach to this problem has been to explore the usefulness of extremely small, lightweight optic flow sensors such as those found in optical mice to determine height above ground.
An optic flow sensor measures the flow of features across a CMOS imaging array. It outputs two values, and , representing the total optic flow across the camera’s field of view in both the x and y directions respectively. The optic flow data in the camera y direction corresponds to lateral motion of the airplane and for purposes of establishing height above ground can be ignored. The optic flow data in the camera x direction can be combined with data from the IMU, GPS, and the field of view of the sensor to determine height above ground. The operation of the optic flow sensor can be conceptualized easily by thinking of driving parallel to a picket fence. If you are traveling at a high velocity and are close to the picket fence then the slats in the fence will appear to be flowing through your field of view very quickly. However, if you are far from the picket fence and driving at the same high velocity the slats will flow through your field of view very slowly relative to the first example. If you know your velocity you can determine your distance from the fence by measuring the rate at which the slats flow through your field of view. For use in UAVs the optic flow sensor measures the flow of features on the ground through the sensor’s imaging array. The velocity of the UAV is determined from GPS data. These measurements combined with the current estimate of the attitude of the UAV are then used to determine the UAV’s height above ground.
One of the most important considerations in outfitting an airplane with an optic flow sensor is the setup of the optics. Narrow angle lenses increase the distance at which the airplane will be able to pick up appreciable optic flow. This increases the altitude at which the sensor will be able to accurately measure height above ground for a given ground speed. This comes at the expense of increasing the longitudinal size of the sensor and causing overflow in the optic flow sensor’s register for low heights above ground. Wide angle lenses allow for smaller longitudinal sensor size, but require that larger features be available and decreases the distance at which appreciable optic flow will be recorded for a given ground speed. Also, larger diameter lenses allow more light into the optic system thereby decreasing the requisite light intensity of the environment. Smaller diameter lenses require an environment with a higher light intensity in the required spectral range for the sensor to operate properly. A main focus of our research has dealt with determining lens arrangements ideally suited for determining height above ground in mini and micro UAVs.
Another important consideration in designing an optic flow sensor system for detection of height above ground is the choice of sample rate. Because the optic flow sensor does not measure height above ground directly, the resolution of the sensor is determined by the transformation from optic flow to height above ground. This transformation is highly sensitive to the rate at which the sensor is sampled. If the sensor is sampled too quickly the resolution of the sensor can quickly deteriorate to values of greater than 20 meters of height above ground per sensor count. This is undesirable not only because it makes it impossible to determine height above ground to within a value that is less than the resolution of the sensor, but also because it amplifies the noise of the sensor to the point where the sensor reading is useless. On the other hand, if the sensor is sampled too slowly the sensor’s register will overflow causing the sensor to give inaccurate readings. The issue of sample rate is further complicated by the fact that the ideal sample rate is constantly changing as a function of height above ground and ground speed. A main focus of our research has dealt with developing algorithms for dynamically shifting the sample rate to maintain one to one resolution between sensor counts and height above ground in meters. Our algorithms have demonstrated the ability to greatly reduce sensor noise at large heights above ground and to avoid overflow at low heights above ground.
Further research efforts will focus on applying current algorithms and hardware arrangements for detecting height above ground to detecting and avoiding obstacles. This will be particularly useful in canyon following applications.
