How to navigate robot wirelessly by mapping?

Hi Sachin,

Mapping requires recording the geometry of the workspace - measured dimensions and locations of all terrain features and objects.

Automating the this mapping exercise requires a lot of infrastructure that only gets used once, so it is often more cost effective to manually map the target area for smaller projects like this.

The system needs to continuously know where the mobile robot is in the map, so this part of the system infrastructure is unavoidable.

A common method to achieve locallization knowledge is to use beacons - either beacons located in known locations on the map that the robot can use to determine its position by triangulation or a beacon on the robot that gets sensed by several fixed location receivers, again using triangulation to determine where the beacon is. This later method could use a host computer to do all the calculations and remote control. The receivers could be wireless, allowing easy placement. For best accuracy, the receivers would need to always have a direct line of sight to the robot, so angles to the beacon may be accurately determined.This may require more strategically placed receivers to get full coverage of the workspace. Beacons could be as simple as an infrared LED, flashing at a unique frequency.

Doug

Doug can you elaborate on the actual hardware that you would choose to create a beacon system if you wanted a robot in the field to be able to retrace the exact same path months or years later as a result of beacon placement?

It sounds like you have some experience with this!

Adam

Hi Adam,

The type of system you choose depends on factors like what type of terrain is present, what type of obstacles are present, how accurate it needs to be, how fast it needs to be, how much it costs to implement and how much time is available for development.

If you always have direct line of sight to 2 or 3 beacons, you can use either angle or distance measurement to triangulate position.

To measure angle requires either a rotating sensor or a sensor array of at least 3 sensors. Each beacon would need to emit a unique frequency, or they could be sequenced in a fixed sequence. For example each beacon could detect its predecessor before emitting its signal.

To detect distance requires accurate timing to determine how long the signal took to get from the beacon to the robot. Some systems use a radio pulse to indicate when an ultrasonic pulse is initiated, so the time the sound takes to travel to the robot is the difference between when the radio signal is received and when the ultrasonic pulse is received.

Distance can also be determined just using ultra wideband radio signals as well, but time measurement must be much more accurate for this.

I have a robot that navigates buildings just using an optical proximity sensor to detect walls, doorways and corners. It doesn't need to know the building floor plan, but that knowledge would eliminate unnecessary exploration.

Another method is to place recognizable patches at fixed way points on the robot path and have an on-board inertial guidance system that is accurate enough to get from one way point to the next without any external signals. At each way point the inertial system can be recalibrated so the long term drift of such systems does not become too large. For example a patch could be just a big Y. The robot could have a line-following sensor system that would kick in whenever a line on the Y was crossed. The V section of the Y  would then  "funnel" the line following robot to its center and the tail of the Y would send the robot in the direction of the next way point.

Here are a couple of links to get you thinking about different methods:

http://www.cs.cmu.edu/~cyberscout/publications/FSR99.pdf

DecaWave RTLS and ranging demo - YouTube

Doug

Thank you so much for the response and I will be studying the links and the information you have provided. I know that this will come in helpful and I may have some follow up questions.

We have a "pavement scanner" yes it is what it sounds like but it has its purpose.

I was thinking of setting up a beacon system over a large line of site area. This would be on a highway. We set female pins that the beacons can be screwed in to.

The hope would be that year after year we can return and scan the EXACT same piece of pavement. It looks like the systems you pointed me towards can help me build that.

In many ways im running a 2 Axis cnc machine.

Can G code help me create a model space and save "tool paths" if you will.

Hi Adam,

If you take a look at some of the IRobot autonomous sweepers, they use a couple of IR beacons to setup cleaning areas.

They use an IR LED set to a fixed frequency, about 40 KHz.  The robot can then determine when it reaches the edge of detection and changes is direction away from the edge.

You can set up this type of beacon very easily with just a couple of photo detectors and some timers to gage you position from each beacon.

You can also use multiple beacons to map out an area.

If you have good positioning data for your robot, you can also setup a simple bit map to layout a preprogrammed course for you robot.

Lots of options exist.

DAB

Hi Adam,

If you are basically scanning in straight lines or a raster scan, you could use a laser to create a line that the robot would follow. (I am assuming you can't just paint a line on the pavement - if you could, it would simplify everything - for example it could be invisible paint detected by UV sensors) Leveling lasers you find in hardware stores often can generate a laser line on a target surface, but for use outside, the laser should be modulated at a unique frequency, so it can be distinguished from sunlight and other light sources. For long range work a line laser requires a fairly high power laser and precision optics.

Line following robots are so simple, they don't actually need a CPU, just some suitable control circuitry. However if the path is a full raster scan a CPU will likely be useful. A robot could also be setup to follow a guard rail at a specific offset distance to one side, with different offsets for each raster scan line. This just needs a suitable proximity sensor.

Doug

Hi Adam,

If you are basically scanning in straight lines or a raster scan, you could use a laser to create a line that the robot would follow. (I am assuming you can't just paint a line on the pavement - if you could, it would simplify everything - for example it could be invisible paint detected by UV sensors) Leveling lasers you find in hardware stores often can generate a laser line on a target surface, but for use outside, the laser should be modulated at a unique frequency, so it can be distinguished from sunlight and other light sources. For long range work a line laser requires a fairly high power laser and precision optics.

Line following robots are so simple, they don't actually need a CPU, just some suitable control circuitry. However if the path is a full raster scan a CPU will likely be useful. A robot could also be setup to follow a guard rail at a specific offset distance to one side, with different offsets for each raster scan line. This just needs a suitable proximity sensor.

Doug