Saturday, August 31, 2013

Robot Explorer Log 7 Stall Sensor

Mercury tilt switch makes a 1-axis accelerometer

The robot build took on more expansions, from 2 drive wheels to 4, the chassis was extended, sensors were restructured, and getting stuck either on the ice or in the marsh of primordial mud soup on other planets and moons was considered. This has led to the development of the stall sensor.
Parallax wheel has ellipsoid holes for encoding

The question initially put forth on the Parallax Forum here has garnered many thoughtful ideas and very interesting responses. We've supplemented the information with some of our own, and a short analysis about what can work and what can go wrong - Murphy's Law and Yhprum's Law.

The use of a high mounted horizontal light measuring sensor could detect if the robot is still capable of moving left to right. This may not work on planets with little or no visibility due to atmospheric pollution and various "white out" storms. For example, some planets have dust storms lasting many months while other moons have thick pollution consisting of a mix with visually blinding methane and hydrocarbons. The adjustable and programmable light sensor which can set sensitivity requires three pins.

Also considered is the use of a free spinning idler wheel which has encoding and contacts the ground. If the robot is moving and not stalled, the wheel will turn as it moves along on the ground. The new Boe-Bot wheels have eight holes which can be used to construct a home-made encoder. The encoder consists of a simple IR transmitter receiver pair, one on each side of the wheel. The encoder, because it's made up of a light sensor, could by happenstance end up useless if it becomes covered in mud, dirt, dust or crud. The encoder takes up two pins, one for the transmitter and one for the receiver.

A two axis accelerometer could determine the forces of the robot starting and accelerating forward along one axis, and confirm left to right movement along another axis, which would not appear if the robot was held in mud. The accelerometer could also have other functions like measuring terrain slopes and inclines, and preventing robot tipping due to tilt. Since these accelerometers work well in sensitive seismometers, they will undoubtedly work for this application. The accelerometer can perform within an enclosed environmental capsule to maintain proper functioning of its built in heater. The accelerometer takes up two pins.

Perhaps the cheapest and most simple approach is to use a side to side (or forward/reverse...) sloshing mercury switch mounted on the robot. This acts as a cheap one axis accelerometer, and can determine if the robot craft is still capable of quick accelerations, i.e. movements left to right, thus proving "it's not stuck" with a returned signal. This would take up only one pin and a small program can poll the status during a short test. This sensor is not effected by dust, atmosphere, or being covered in mud. It can also reside within a temperature controlled enclosure, not exposed to the elements outside. Keep in mind, mercury freezes at -38 degrees F. and boils at +674.11 degrees F. This agrees well with other sensors with a -40 deg. F. operating temperature limit. The mercury switch is a cheap, inexpensive, poor man's accelerometer and takes up only one pin.

Ok, so now we know the robot is stalled. What can we do?

The "Unstuck" program will help take care of that. The program is designed to alternate the drive wheels back and forth to generate momentum that should free the robot. This is similar to a stuck car when the driver can alternate putting the car in reverse and forward drive, rocking the vehicle back and forth, and building up enough momentum to break free.

Still stuck? If the planet has changing weather, such as changes in surface conditions, it may be possible to wait and utilize these conditions to better facilitate the "pulling free" process. The unstuck program will be tailor made to the conditions on the planet or moon. Conditions may include a cycle of freezing and thawing, a dust storm, particle rain, wind, temperature changes and other effects.

The Unstuck program requires no pins, however, it will make use of several sensors which require pins. The Momentum Driver remains in software, making use of four wheels and timing.

There are some fantasy rescue ideas too:

A tethered quadcopter takes to the air and helps pull the robot free from the mud.

The robot, using spring loaded underside flaps, jumps out of the mud.

A rescue shadow Boe-Bot flings forth a grappling hook and wenches the robot free.

A 10-foot rope is anchored to the ground. After Boe-Bot advances 10-feet, the rope/anchor connection is auto-released and the rope is retracted. The robot anchors the rope again, advances another 10-feet and the process is repeated. Should the robot become stuck, the rescue rope is rewound, helping to pull the robot to safety.

As the planet may have no magnetic field, a compass cannot be deployed alone.

Electromagnets will work in space and on planets without a magnetic field. The Boe-Bot drops a small tethered weighted electromagnet from the front, to the ground. As the robot moves forward over it, the tether unwinds and the rear hall effect sensor detects the magnet, proving the robot has moved and is not stuck. The small electromagnet is retracted and the process repeats.

If Boe-Bot is on a wind planet, deploying a sail and waiting for wind to help pull free could be a solution. Aligning the sail's force vector with the inline direction of wheel travel could be helpful. Choosing a forward or backward direction for turning wheels can be useful and offers more options. A wind sail program can code the wheels to drive in the direction of the wind sail's force vector. This may need a wind direction sensor. However, a static wind sail will push the vehicle either forward or reverse. Alternating the driving momentum, both forward and aft, can make a wind direction sensor unnecessary.

Parallax Forum 
The Element Mercury
Yhprum's Law
Propeller C Activity Bot