Programming & Simulation Lecture 6

Loading...

En provenance du cours de Georgia Institute of Technology

Control of Mobile Robots

1103 notes

Georgia Institute of Technology

Control of Mobile Robots

1103 notes

Control of Mobile Robots is a course that focuses on the application of modern control theory to the problem of making robots move around in safe and effective ways. The structure of this class is somewhat unusual since it involves many moving parts - to do robotics right, one has to go from basic theory all the way to an actual robot moving around in the real world, which is the challenge we have set out to address through the different pieces in the course.

À partir de la leçon

The Navigation Problem

Behaviors Revisited13:37

Hard Switches vs Blending10:38

Convex and Non-Convex Worlds9:25

Boundary Following9:35

The Induced Mode8:57

A Complete Navigation System9:27

Practical Considerations9:04

Let's Do it!4:39

Glue Lecture 68:51

Programming & Simulation Lecture 612:49

Rencontrer les enseignants

Dr. Magnus Egerstedt
Professor
School of Electrical and Computer Engineering

Welcome to week six.

This week I'm going to be talking about following walls.

Why do we need to follow walls?

Well, last week we solved the navi,

we somehow seemed to solve the navigation problem.

We had our robot drive from point A to

point B without colliding with any of the obstacles.

But, the obstacles were fairly easy, and.

The reason for this is that they're convex.

But we can have these obstacles that are concave in shape.

So, for example,

take this U-shaped obstacle, and if a robot starts here

and it's goal location is on the other side of

the obstacle, what will happen with the controllers that we

designed last week is that the robot will drive from here.

Into the obstacle, and then start avoiding the

obstacle, and then drive back into the obstacle.

Because it's always trying to get to this side, but also avoiding this obstacle.

But it's trapped in here.

because the shape of the obstacle is concave.

So what we need is an additional controller, the follow-all

controller, that will, so that when we make a decision

that we're inside of our concave obstacle, then we're going to

follow the contour of the, the boundary of this object, and.

We're going to follow it all the way around until we've

cleared it and then we can proceed onto the goal location.

And that's, that's the objective for this week.

And the way that we'll do this is

we're going to use the infrared sensors to approximate

a section of the wall, alright?

It's really an obstacle but we're going to call it a wall.

And.

And once we've, once we've estimated a section, we're basically

going to figure out a vector that's tangential to the wall.

And also one that's perpendicular to the wall.

And the point is to combine these two vectors into a single

vector that will not only steer the robot into,into one direction along.

The the, the, the contour of the, the boundary of the wall.

But also a vector that will make sure that the robot

neither steers into, into the wall or away from the wall.

So we're going to maintain some distance from the wall as we follow it.

Now.

How is the robot going to estimate the wall?

Well, first of all, one of the two inputs into the

controller is going, is going to be either left or right.

So we are either going to follow a wall thats to our left or to our right.

So we have to make that decision before we do anything else.

If we decide that we are going to use.

The left the left direction. What we're going to

do, is we're going to use sensors one, two, and three on the robot.

So we're going to use these three sensors.

to, to estimate a wall that's the left, or

an obstacle that's to the left of of the robot.

And.

What we'll do is we'll pick of these three we're

going to pick the two, that have the shortest infrared sensor readings.

And, we already know how to get a point, on these in the world frame.

So, what we end up is we end up with a point so.

We have, this point right here which I'm going to denote as, point 1.

And, this is going to be point 2.

So, we've picked these two, and since we've picked two points

in the world, we can come up with a vector right?

So, we're going to come up with this green vector right here.

And, this green vector is going to be our vector

that's tangential to the wall.

So I'm going to call this u, the vector u

fw denotes that, this is for the follow wall, and.

Comma T denotes that this is tangential to the wall.

And really, the way that you define it is, it's p2 minus p1.

That's the definition of this, of this vector.

Now, I also said that we want to maintain some spacing from the wall.

So we're going to have to calculate

a vector that's perpendicular.

And the way that we're going to calculate it is,

again, we have the vector from the previous slide.

We have u_fw_tp, and this time it has a prime

mark, and that's because I'm going to normali, normalize it, and.

Thus, and again, I think last week I talked about this is, what we're

going to do is we're going to just take

the vector, and divide it by its magnitude.

So that's the,

that's the normalized version of the of that vector.

And what we'll do is we'll first take a

point on this vector, which is going to be ua, and.

This point, I'm just going to pick for, to make it easy, I'm going to pick the first

I'm going to pick the point detected by this first infrared infrared sensor.

So, this one right here.

because I know this one I could also picked.

This one right here but I picked the first one.

Now I'm also going to pick another point which is

going to be the location of the robot in the world.

So that's going to be my point or vector up, u sub p.

And since I again have but I have so, I have this point so I have.

Let me indicate that on the slide. I have

this point right here, this point right here, and this vector, and what I can do

is I can compute the, a vector that points from the robot, so from the robot.

To this closest point on the wall.

And the, and the equation that describes this vector is as follows.

It's this, so this vector, this perpendicular

vector I'm denoting by u sub fw, p.

And it's this subtraction of these points dotted.

With a dot product of, of the tangential vector

and then multiply it again by the tangential vector.

So this is just some linear algebra that gives me a vector from the

robot to the closest point on the

wall, that's, that lies on the tangential vector.

And, this is not sufficient for us to maintain spacing yet.

What we really want to do, is we want to find another vector that's goes in

the opposite direction, and, and again from

the previous slide I have this perpendicular vector,

so all I'm going to do is I'm going

to normalize it and multiply it by some distance fw.

So d sub fw.

And.

This distance right here is the distance that

we want to maintain between the robot and the walls.

So we want the robot to end up in such a way that this is equal to dfw.

So the reason that I define this vector like, like so is that when we can combine.

The two together.

If we combine these two vectors together, we're

going to get a vector that either points us

toward the wall if the robot is farther than

d fw from the obstacle, or from the wall.

And we're going to get a vector that points us.

Away from

the, from the wall if we're closer than d sub fw.

So, that's the, that's the purpose of going through, through these equations.

And really, what ends up happening is if we div if we.

Somehow linearly combine the, tangential vector to the, to the wall.

And the perpendicular vector

the, to the wall, that you know either

points us towards or away depending on the distance.

If we combine those in some way, we're going to get a, vector,

u sub fw, which will both.

Point us in the direction in which the wall is going to, that allows

us to follow the wall, and it also has the component that either pushes us towards or

away from the wall depending on how close or far we

are away from the obstacle, and this is exactly what we wanted.

And then the last thing that we have to do is we just have to use our good

old PID controller to steer the robot in the

direction of this vector, and we're good to go.

So how all does this work?

Well, you saw in the previous slides that I got a really good

estimate of, of, of the, of the, of a section of the wall.

But that's not always the case.

So, for example here on the inside corner, what I have here is

this is the robot's estimate of, of the, of the wall.

And you can see that this is an overestimate.

So, we're actually okay with that, because if

we're overestimating the wall, that's, you know, that's

not really problematic because we're not going to.

We're going to avoid this, this virtual piece of

the wall and we're not going to slam into it.

There's no danger of, real danger of that. But that is on, these inside corners.

On the outside corners we are underestimating the wall.

So you can see here that the robot thinks that this represents.

Is it

a good representation of the wall?

And if we cut this corner too closely the robot collide with this

part of the obstacle that it doesn't really sense because there's a little

bit of spread between the sensors because we don't have, you know, 20,

20 sensors on the robot that cover every single inch outside the robot.

We only have five sensors, so we got to make,

deal with what we have, so here we have to

be a little careful, and that's where the design of this distance, d

sub fw comes into play. So, you want to make sure that this, this,

value is large enough to allow us to not cut corners and collide with the wall.

But also not too large so, so that we don't lose track

of the wall, because we don't want to end up too far away

from the wall and then the robot doesn't sense

the wall anymore and we can no longer follow it.

So, we want to kind of stick to the wall as

close as possible, but not too close to collide.

Now the implementation is going to happen in a new controller, FollowWall.m.

And you're going to have to implement some missing logic and,

and, and math that was covered in the previous slides.

And ag, also again covered in

the, in the manual in much greater detail. And.

As I said before the, what's, what's really unique about

this controller its going to accept either left or right

denoting which side of, which side of the robot the

wall is going to be in which its going to follow.

So we're going to tell it you're going to have to

follow a wall that's going be on your left

or we're going to tell you you're going to have to

follow an op a wall that's on your right.

And that those are going to be inputs and those [INAUDIBLE], really come into play.

And next week when we have our full navigation

algorithm and state machine going on.

So, let's see what this looks like in, in MATLAB and in the simulator.

I'll launch the simulator and hit Play.

And here, we have the robot, following the wall on it's left.

And as you can see here I got really, really close, to this corner.

So that was a little dangerous.

Maybe I should have picked a better,

better distance to maintain from, from the wall.

And here we are overestimating so we're okay.

And here we're going to cut it fairly close again.

So this is pretty, pretty much a little too close for comfort.

But, you know, we're not crashing into the, into the obstacles.

So we're, we're, we're gon, we're okay with that.

Alright. Great.

So, the robot's going to follow this wall on and on and on until I close this.

Now, again, make sure re-read the manual

for weeks, that's been updated for week six.

It has a lot more details especially, you know, for

the math and, and has diagrams that will help you.

Implement the logic in all the math in in the, in the controller.

And

I encourage you to experiment with different values

of the distance to maintain from the wall.

So you'll see different results for different distances that you pick.

And also you can decide how to combine the two vectors that

are tangential and, and perpendicular, to the, to the wall, and, and, and.

In different ways, so.

And it depends on where you want to put the emphasis.

You might want to put the emphasis strictly

on following the law, on following the

law, and not worry about maintaining a distance

too much, or you could do the converse of that, it's really up to you.

And with that I wish you good luck.

Explorer notre catalogue

Rejoignez-nous gratuitement et obtenez des recommendations, des mises à jour et des offres personnalisées.

Coursera propose un accès universel à la meilleure formation au monde, en partenariat avec des universités et des organisations du plus haut niveau, pour proposer des cours en ligne.

© 2019 Coursera Inc. Tous droits réservés.

Télécharger dans l'App Store

Disponible sur Google Play