This post reviews “Passive sensor
evaluation for unmanned ground vehicle mud detection” by A.L. Rankin and L.H.
Matthies. The article originally was
published in the Journal of Field Robotics, July 2010, Volume 27, Issue 4. The military is probably the biggest
proponent for developing unmanned ground vehicles. This includes everything from robots which
can disarm bombs to trucks which can drive from point to point without a human
driver. All of these UGVs are going to have
one thing in common; they will be expected to operate in some of the most aster
environments on Earth. One of the things
a human driver in a truck has going for them is the ability to look ahead and
distinguish favorable terrain from that which will cause the vehicle to get
stuck. A UGV is going to need similar
abilities before they could become dependable parts of a military force. There
are a multitude of proprioceptive sensors which can detect things like wheel
slip to indicate the vehicle is already in the mud, but an exteroceptive
sensors are needed to prevent the vehicle from getting into the mud at all. Rankin and Matthies conducted a series of
tests to determine what passive sensors or combination of sensors would work
best for detecting mud ahead of a vehicle.
The
technology to detect the moisture content in an area of land is not new. It has been researched and used in
agricultural applications for years.
This same technology has some applications to UGVs, however, most of it
is not transplantable to a vehicle application which can look forward and make
a distinction fast enough for a fast-moving truck. Rankin and Matthies broke their tests down to
different types of cameras/sensors available for use. This included stereo, color, near infrared
(NIR), short wave infrared (SWIR), mid wave infrared (MWIR), long wave infrared
(LWIR), and polarization. The authors do
a phenomenal job of describing the successes and shortfalls of all different
types of cameras. The biggest conclusion
of the tests was no one sensor is the be all end all solution. Depending on daytime or nighttime operations,
there would be a combination of different sensors which would provide the most
successful results for detecting the wet soil prior to the vehicle becoming
stuck. Overall, the article was very
well written. The information presented
was easy to understand and leaves very little questions to be answered on where
the technology for mud detection needs to go.
References
Rankin, A., & Matthies, L. (2010). Passive
sensor evaluation for unmanned ground vehicle mud detection. Journal of
Field Robotics, 473-490. doi:10.1002/rob.20341
Interesting post. Evaluating moisture content in soil while driving on unimproved roads and trails is something human operators do without much deliberate thought. When the operator is removed from the vehicle, filling all the gaps in perception must be addressed. Determining whether the ground is too moist to effectively transit is one thing that probably hasn't received much attention. Mud can definitely threaten the mobility of an off-road vehicle. I'm interested if Oshkosh Defense has addressed the issue with the TerraMax system.
ReplyDeleteBrandon,
ReplyDeleteReally great post. I found the article interesting and your take on it was right on. There has to be a way for the vehicles to know what type of surface it will be on and or going into. The math that is involved for the algorithms must be crazy to figure out the difference in all the different surfaces and to know exactly at what point the vehicle would get stuck. I guess they need to get the vehicle stuck first to find out what its capabilities and weaknesses are. Great Post.
Brad Simon
Brandon,
ReplyDeleteIn addition to the military, technology such as this will be very useful for driverless cars in the near-future. Although presently most cars with driverless capability are limited to operation only on highways with clear lane-markings, in the future this technology could be effective for off-road driving, such as when traveling to a campsite. Great information!