Research: High Frequency Radar and Robot Perception
Radar holds the prospect for modeling, safeguarding, and navigation
from a space robot with advantages of operating in and through dust, in
vacuum and atmosphere. Size and resolution are major drawbacks for this
technology. Non-mechanical scanning antennas or antennas with simple designs
size. I have researched to identify technologies and vendors that develop
non-mechanical scanning antennas or very simple mechanical ones.
Software narrowing techniques reduce it further by enhancing wide beam
(i.e.. small antenna aperture) sensor data. The adoption of the "evidence
grids" technique promises the use of wide beam sensor data to produce very
detailed maps, without characteristic problems of radar perception (for
example, rejection of side lobes). Another approach is the adoption of
Synthetic Aperture Antenna techniques. Addressing these issues is the agreed
thesis topic with my advisor.
Comprehensive testing of an existing system has been the major activity
of the pre Antarctic weeks. A thorough test to characterize the sensor
was recently completed. The testing included range and angular precision
and resolution measures, mixed pixel phenomena, minimum detectable gap,
beam shape and grazing angles for various soil types. A formal report of
this activity will be available shortly.
Specifically for this Antarctic season, I work to produce a test of
millimeter wave radar to validate it as an adequate perception modality
in all Antarctic conditions. Performance for different grazing angles and
different terrain in Antarctica (blue ice, snow, sastruggi) and atmospheric
penetration under severe visibility conditions are main topics to explore.
Imaging radar research at Carnegie Mellon University explores the use of
this sensing modality for the perception of field robots, both in terrestrial
and planetary applications. Millimeter wave radar is preferred to sense
in environments that are dusty and have changing light conditions.
The sensor size and mechanical complexity challenge the applicability.
The size is driven mainly due to the minimum antenna aperture necessary
to achieve a narrow beam for high angular resolution. The need to scan
a beam to provide the imaging capability requires several moving parts,
compromising the reliability and space worthiness of the sensor.
The current research seeks to answer these shortcomings by adopting
techniques from other perception fields that have wide beam sensors (synthetic
aperture radar, ground penetrating radar, sonar) to achieve comparable
resolution maps using smaller antenna apertures. The mechanical complexity
is addressed by embracing non-mechanical or one-moving-part scanning antennas.
Scanning Antenna Technologies
The research activities will in the identification of the most promising
technologies. The selection criteria include the mechanical simplicity,
a potential low production cost, and the actual existence of demonstration
hardware. The following list includes technologies that are explored.
Optical beam forming uses a photo conductor where the illumination changes
the diffraction properties of the material, thus attenuating incident millimeter
wave radiation. Because the material responds rapidly to changes in the
incident light, it is possible to form a scanning beam without moving parts.
It is possible to construct an antenna capable of 60 degrees of angular
scanning in two axes. The beam width at 94 GHz is 1.6 degrees in both planes.
The antenna might operate at any frequency in the 2 to 100 GHz range. Size:
6 - 11 inches. Fall back: very low efficiency.
Ferrite control changes the magnetic field along a dipole array antenna,
thus changing the phase shift between each consecutive dipole. A simple
current control scans of the beam. There are linear one-dimensional scanning
antennas available for 77 GHz and planar two dimensional antennas for 38
GHz. The 77 GHz antenna scans a range of 15 degrees, and it is possible
to double the figure by double feed. The antenna is non-reciprocal (transmission
and reception patterns are not the same), therefore two are needed for
radar applications. This requires a more complex control strategy to synchronize
the two antennas.
Spinning grating antennas are capable of providing fast linear scan while
requiring only continuos speed rotation. They are characterized for very
low cost and simple design. Current applications are automotive collision
avoidance and aircraft autonomous landing systems. This technology
uses one moving part for one axis scanning.
Software Narrowing Techniques
"Evidence grids" technique uses a learned model of the sensor to weight
the evidence of presence or emptiness. This evidence is integrated into
a grid if cells, so the certainty of emptiness or occupancy grows with
the number of observations. The ongoing development of the radar beam model
pursues the application of this technique to radar beams. The technique
has the potential to sharpen the images and correct for erroneous side
lobe information. The field experiments with the FMCW imaging radar will
provide the information to build the sensor model and apply this technique
to real scenes. The rejection of side lobe ghost targets is a major expected
improvement of the application of this technique.
Synthetic aperture radar techniques use the relative motion of the sensor
with respect to the target to synthesize a larger antenna. Field robots
are generally mobile and that feature shows promise to increase the accuracy
of the terrain maps.
The first stage of radar testing finished recently. The selected area has
no undesired reflections and enough area to lay targets and vary the geometry
of the scene. The completed phase includes the following tests:
An early conclusion of this experimental activity is the need for better
range resolution. A market research for state of art technology in short
range, high-resolution radar transceivers is ongoing.
Range precision (sensitivity to small changes of the target's range)
Range resolution (ability to resolve two targets in the same angle)
Angular precision (sensitivity to small changes of the target azimuth)
Angular resolution (ability to resolve two targets at the same range)
Beam shape (energy returned from a target in different position with respect
to a stationary beam)
Minimum detectable gap (between to surfaces perpendicular to the sensor)
Range boundary (mixed pixels where the range of the object changes drastically)
Side lobe characterization (ghost targets that appear to be in front of
Grazing angle performance (minimum incidence angle that results in the
surface detection by the radar).
Current Sensor Description
The radar imaging research is undergoing an intensive experimental phase.
A 77 GHz FMCW imaging radar was made available for this research. The
sensor scans horizontally four vertically stacked beams. Each beam
is one degree wide and two degree high. The horizontal scan range is 64
degree and the four beams cover 8 degree in elevation. The angular range
is mechanically scanned at 2.4 times per second and the angle increment
is one degree. The radar uses FMCW (Frequency Modulated, Constant Wave)
range estimation with a maximum range of 64 meter and 0.5 meter range increments.
Plans for Antarctica
In November 1998, a MMW imaging radar will be deployed at Patriot Hills,
West Antarctica. MMW radar is a preferred imaging sensor modality
because it provides precise range measurements for the environmental imaging
needed to perform autonomous operations in dusty, foggy, falling snow occluded
(white out, blizzard) and poorly lit environments.
The backscatter test will measure the energy reflected in a direction
opposite to the incident wave. This test will explore the energy returned
at different grazing angles and on different surface types (blue ice, flat
ice and snow). These results will be compared to the performance on a similar
geometry of regular dirt. This data will help to build a database
of performance of MMW imaging radar on different terrain.
The atmospheric penetration test will measure the degradation of performance
of the sensor under severe visibility conditions. The same scene will be
sensed under clear air conditions and also under heavy flying snow (whiteout,
blizzard). Visual conditions will be recorded by photography of the scene.
A laser scanner will scan the same image for comparison purposes. A quantitative
estimation of performance in severe polar vision conditions will support
the application of this sensing modality.
Burying metallic targets and scanning with a frontal beam will test
solid snow penetration estimation. The targets will have increasing depths
up to a few centimeters. Only shallow penetration is expected.
Short Term Plans during and after Antarctic Deployment
out about the millimeter wave radar experiment in the current Antarctic
Select the best scanning antenna technology and characterize the beam pattern.
Assemble the antenna to an integrated millimeter wave transceiver with
high range resolution.
Demonstrate functionality of software beam narrowing techniques. Specifically,
apply the "evidence grids" principle to solve the side lobe issue. Establish
metrics to relate application requirement with maximum beam width tolerable
with the software narrowing techniques.
Determine the applicability of synthetic aperture radar to field mobile
robots. Specify requirements of position estimation that would enable the
utilization of this technique.
Robotic Search for Antarctic Meteorites 1998
All material on this page is property of NASA and Carnegie Mellon
University. Any image or text
taken from this site and incorporated into another document without
consent violates the Copyright
Law of the United States and the Berne International Copyright
Send comments, questions, or suggestions to Dimitrios
This document prepared by Michael