Steering

So what we are doing is an accelerated evolution of sailing technology.

Can you steer a course?

It'll make a great tail.

We went from a back rudder to a front rudder to two rudders to multiple rudders to the whole boat changing shape, and the more we are moving forward, and the more the design looks simple and cute.

Out of sight.

My turning angle, I called turn.

The wing has no steering controls, no flaps, no rudder.

Let's assume the car has a fixed forward velocity, but you have the ability to set the steering angle of the car.

Okay. Thank you. Would you Like to take her a while?

Here is the problem.

I should warn you that sometimes I do get collisions here, and it's because our obstacle surfaces are relatively large.

The vehicle is actually modeled as a system with 2 non steerable wheels and 2 steerable wheels.

R times the sine of theta, and cy is equal to the initial y position plus this extra distance, R times the cosine of theta.

It then reduces y over time to go into negative territory.

Looking at him, I could see white sails and blue water.

I'm leaving you in command. Get the ship operational.

If I set the mean of the Gaussian to be the steering command and the distance command then this adds noise.

You can see the steering tends to be on the negative side.

The crosstrack_error by and large is very small.

During the day, the Nutmeg collects energy from solar panels on the roof of the wheelhouse, and from a 450 watt turbine up the mast.

We manipulate this equation a little bit, and we find that the radius of curvature is equal to the length of the vehicle over the tangent of the steering angle.

Here I'll apply the noise, which you don't need it for the first implementation, but later on as we go on, you need it to actually add noise.

Suppose you do what I just said. You steer in proportion to the crosstrack error.

So you start steering more and more the more time goes by to the right to compensate this bias.

That's measured by the integral or the sum of the crosstrack errors over time.

But let's try this out with out our proportional controller.

Put differently, the robot oscillates a bit like this with a fairly constant new offset error due to this bias.

That is, it's not just points in x y space, but it's successive state vectors for cars.

I make a new robot copy as in my sample code in class.

Then I was so happy that I kept playing with the robot, and so you see the robot has a front rudder here.

A constant steering will put you in a circle and not in a straight line.

Then we apply the corresponding new motion model to x and y to obtain x2 and y2.

I just add Gaussian noise with the corresponding steering noise and distance noise parameters.

What's the intuition?

You have to fight for what you believe.

You will reach the trajectory. Clearly the other two answers are really bad.

As a result, when you drive you steer the car this way.

The second part we didn't talk much about is selection of speed.

How would you do this? You would keep the steering constant? You would use random steering commands?

A normal sailing boat has a centerboard here and a rudder at the back, and these two things are what creates most resistance and turbulence behind the boat, but because this doesn't have either a centerboard or a rudder, we hope that if we keep working on this hull design we can improve and have less resistance.

Then I started publishing online, and some friends from Korea, they started being interested in this, and we made a boat which has a front rudder and a back rudder, so we started interacting with this, and it was slightly better, although it was very small and a bit off balance, but then we thought, what if we have more than two points of control?

With this night's revels and expire the term Of a despised life, clos'd in my breast,

The radius of curvature is equal to the length of the vehicle over the tangent of alpha where alpha is our steering angle.

And so, I thought, what about if we just take the rudder from the back of the boat to the front, would we have better control?