Robotics question

[OP No.7470139]

I'm building a robot and I've run into a dead end with regards to sensors/guidance. Any ideas/semi-productive shitposting would be useful.

I am designing a robot that needs to be able to drive from a starting point to another point in a *perfectly straight* line for 100m (330 ft). For simplicity's sake let's define that as <1 cm error. What is the cheapest way to do this?

If necessary you can mount equipment on a pole at the destination point. But keep in mind that some robots could have dozens of separate locations that they have to perform this "straight line task" at. So emphasis on the cheap.
Also, feel free to include a one-time calibration phase to acquaint the robot with it's environment(s).


List of rejected ideas will follow in the second post (for those so interested).

[OP No.7470141]

Forgot to mention: ground will completely flat. Dirt/gravel is the most inhospitable terrain the robot will encounter.

[OP No.7470157]

>>7470139
Current idea:
Mount a laser pointer (relatively bright even in sunshine) above the destination point, have the robot have three photoreceptors in front of it.
If the left photoreceptor detects the laser, make a slight course adjustment to the right.
If the right photoreceptor detects the laser, make a slight course adjustment to the left.
If the center detects the laser, do nothing.
Would have to have a very fast refresh rate (many times a second) to avoid going excessively off course.


Rejected ideas:
Laser beam homing: was going to have robot home in on LIDAR, but this is also prohibitively expensive. Might be feasible if they were less than or equal to $1000, but most are in the $4-5000 range.

Compass: too much margain of error. Even very expensive compasses are ±0.1 degree, which at 100m results in >1cm error (several inches IIRC).

GPS: not accurate enough.

Cell phone: robot may be in locations without cell coverage.

Magnetic detector (i.e. robot follows a buried wire or something similar): this is plan C. Expensive and only ideal if literally nothing else will suffice.

Might be some others but I can't recall

[OP No.7470192]

Anyone?

[Anonymous No.7470314]

pls respond

[Anonymous No.7470332]

Is there a dead line for this? I will think of something. Just for fun, take a look at this thread >>7466644

[Anonymous No.7470361]

>>7470332
Wtf is this?

[Anonymous No.7470589]

>>7470139
>> *perfectly straight* line for 100m (330 ft). For simplicity's sake let's define that as <1 cm error.
any way you slice it, that's gonna be really fucking hard.

Even with SLAM that's gonna be difficult

[Anonymous No.7470597]

>>7470157
GPS relative accuracy is greater than absolute, are you sure yoy understand commercial gps?

[Anonymous No.7471269]

Paint a line on the ground and use two sensors looking down.
What sensors you use would depend on the setting and what color line you use, but this is what research and testing is for. Can't really help you there, since we don't know what the exact situation is. Perhaps a purple line with color sensors?
You can probably find a line-painter at a sports goods store. They're really just a stick with a wheel attached, but you get one of those with some (environmentally safe) spray paint.
Cheap and easy. IF you can prepare a track of course. If you can't, use something optical to look for a pole. It's flat, right? You could just shoot a beam (IR or whatever) from the pole between two receivers on the robot. Or just one receiver if you're clever with programming.

[Anonymous No.7471283]

>>7470139
I'm sure this is a solved problem in engineering.

[Anonymous No.7471328]

>>7470597
Can you elaborate?

My understanding is that it was accurate to within a few centimeters. But still above my desired accuracy?

[Anonymous No.7471340]

>>7471328
I'm just pointing out that in GPS, there's a significant difference between absolute position and relative position.

The relative positioning of GPS can be very tight, so it may not be written off so easily..

[Anonymous No.7471347]

>>7470157
The one at my university used a bunch of sonar sensors. It had 8, iirc. If the ones on the left measure a decreasing distance, that means the robot is drifting left, so you turn right a little. If the ones on the right measure a decreasing distance, its drifting right, so you turn left a little.

It wasn't very precise, and the stupidest shit would screw it up, but it worked ok for navigating a straight hallway.

You're simply not going to get a perfectly straight line within 1cm margin of error, especially on anything by perfectly flat, smooth floor, on an amateur's budget.

[Anonymous No.7471364]

>>7471269
I have done the line on the ground thing and also the beacon (using IR) with a very simple circuit, just a pair of phototransistors really, but unless you implement a PID controller or go very slowly, you tend to get oscillations that quickly knock you off course. It's inherently unstable for a wide range of realistic parameters.

[Anonymous No.7471372]

>>7471283
Making robots go in straight lines in any environment just from the sensors on it is actually a semi-hard problem.

[Anonymous No.7471377]

>>7470139
just cheat and put a gyroscope on the machine.

[Anonymous No.7471426]

what about combining two GPS trancievers or three on it. make them in a 10cm per leg triangle on the top, have them correlate the data received and calculate true position based off of that and data from a triad of dipole mercury tilt switches that have them angled one per axis... the intertia should activate the corresponding switch pole to the opposite of the direction of motion, making an inertial guidance unit. or harvest the six axis unit out of a PS4 controller

tl;dr
inertial guidance, cross referenced to 3 GPS trancievers to calculate true position.

[Anonymous No.7471502]

>>7470139
>>7470157
>>7470192
>>7470314
You need our halp?

You might want to check over at >>>/g/.
This board is pretty dead.

[Anonymous No.7471513]

>>7471364
what're you using for a control system? id imagine that it would be stable if you shove a microcontroller in there and do the control in software properly

[Anonymous No.7471595]

>>7471377
This.