The water slide video belong to the Discovery Channel or Beyond Productions. All I can do is post links to the videos that they have seen fit to leave on the internet. The full original program is no longer available.

I was not there. One of our engineers Dennis Ritola was there for many days programming and tuning the systems. Dennis is visible with his Delta Computer Systems shirt on in the final video where Joe Grand takes the first ride.

The mechanical parts were largely built by people that hope to get more visibility or advertising. In the program you can see the Prototype this team do work while the cameras were on but their part was really minimal.

Our distributor
http://www.hydraulic-controls.com/
didn't even get mentioned in the credits even though they supplied our controller and 6 Bosch servo valves and actuators. The design was not done by our distributor. It was done by someone clueless about hydraulic servo systems. I could never find out who was responsible for such a poorly designed system.

Computer information was supplied by a company called that makes water slides. I believe the company name is Splashtastic. Obviously the computer information was for a virtual waterside because it would be expensive to build a water slide that big or long. Zoz, the blond haired guy with the Mohawk, convert the data to positions and accelerations that the water slide must move to. This requires a very good understanding of physics. The position for the water slide were downloaded to the motion controller as a function of time for each of the actuators. I don't remember how many points there were per actuator but the water slide program took about 3.5 minutes to run. That would be about 2000 points if a point there was a position specified for every 0.1 seconds. There are 6 actuators so that is about 12000 points. That is just a small fraction of the 200000 that we can store. When the start command given all 6 actuators were executing their own curve or spline synchronously.

The programming was easy. The tuning was not. The long hose lowered the natural frequency of the system significantly. It also takes time for the pressure waves to travel the 50 ft. Pressure waves travel about 4 ft per millisecond in hydraulic oil. That is about 12 millisecond delay or dead time which is an eternity in motion controller terms. Compensating for the dead time can be done by just lowering the PID gains but compensating for the low natural frequency requires tricks that most motion controllers can't do.

The hydraulic water slide cannot be controlled by a simple PID controller. Feed forwards are required to predict the open loop control output. Our controller has a velocity, acceleration and jerk feed forward. Normal motion controller only have a velocity and acceleration feed forward but hydraulic controllers require a jerk feed forward. Likewise the PID can't be a simple PID. A second derivative term is required. The derivative gain is multiplied by the error in velocity. The second derivative gain is multiplied by the error in acceleration. The second derivative gain is critical for hydraulic systems with low natural frequencies. It makes the difference between success and failure.

Using a second derivative gain is new. Even we don't have much experience with tuning these systems. We know the mathematical theory but we use our software as a fall back tool. If you look closely at the videos I posted you can see on the graphs a MSE or mean square error. One can tune a system by adjusting the gains so that the mean square error is minimized. If a gain adjustment causes the mean square error to increase then you know the gain was adjusted incorrectly. If the mean squared error is reduced then you know that the gain was adjusted in the right direction.

In the past I have always told my customer to design hydraulic systems with high natural frequencies. The servo valve should be placed on top of the cylinder. If they didn't do this we couldn't help. Our new controller can compensate for some of the poor designs and those systems with a low natural frequency but only to a point. To compensate for the low natural frequency the control output must be exaggerated to compensate for the oil that swells the hose. This works to a point but eventually the +/- 10 volt limit is reached. The water slide motion profile had to be have some of the accelerations reduced so that the control signal wouldn't go beyond +/- 10 volts.

http://www.google.com/search?c...de&ie=UTF-8&oe=UTF-8

http://www.flickr.com/photos/j...s/72157608888589861/

http://www.trendhunter.com/tre...-from-prototype-this

The Prototype this team had one failure. It was a 6 legged walker.
We could have made that work.
http://www.fancast.com/tv/Prot...egged-Vehicle/videos

Our controller could synchronize the legs. It is also obvious that each leg must be able to lift half the weight.

quote:

Originally posted by Peter Nachtwey:


The Prototype this team had one failure. It was a 6 legged walker.
We could have made that work.
http://www.fancast.com/tv/Prot...egged-Vehicle/videos

Our controller could synchronize the legs. It is also obvious that each leg must be able to lift half the weight.

I've seen Boston Dynamics RHex ...( the model used in the clip as the basis for the full sized prototype ) ..in my Robot Wars related travels , it's not in my opinion a "walker".

Boston Dynamics made RHex and BigDog which IS a walker.

http://www.youtube.com/user/Bo...cs#p/u/8/3gi6Ohnp9x8


http://www.youtube.com/user/BostonDynamics#p/u


Regards Woody