Fuzsy

Member
  • Content count

    12
  • Joined

  • Last visited

Community Reputation

136 Neutral

About Fuzsy

  • Rank
    Member
  1. Thanks Emergent, I am currently learning a lot from replies, links and books I got here at the forum. I have managed to get something working but there is still a long way to go. My current solution is based on Hierarchical A*. My goal is to be able make a motion plan in less than 1 second. At the moment I am down to a average of 450 ms, but out of 1400 test positions 50 fails to find a plan. At least within 40 seconds. 40 seconds equal approximately 6000 branches searched. I will look in to RRT. It seems like something I could use. Thanks again.
  2. Thanks Jeffery and Snowman, To Jeffery: I’ll got some ideas on how to optimize collision detection, but if it don’t works I will probably post it as you suggested. To Snowman: It looks like a very good and well written book. I think I will gain quite a lot by reading it.
  3. Hello again, Jefferytitan, your idea with A* as a planner worked. The system is now able to calculate a movement path. But.. yeah.. there is always a but.. It is very slow.. it takes about 15 seconds to calculate. It is the simulation of the movement which determines if a movement will result in a collision which takes time, so I have to look in to some faster collision detection, but that is a whole other problem. Thanks again.
  4. Recommended Ai Books And Sites

    [b]Encog - Framework[/b] A very nice open source framework with good examples. The framework + source code can be downloaded for Java and C# The framework contains a lot of different kind of neural networks and learning techniques such as: Feedforward Neural Network, Boltzmann Machine, Hopfield Neural Network, Genetic Algorithm Training, Backpropagation, ADALINE Training etc. A lot of examples, video lectures and a wiki about networks and learning techniques can also be found on the homepage. [url="http://www.heatonresearch.com/download"]http://www.heatonresearch.com/download[/url] [b]Aforge.NET - Framework[/b] The framework contains the most basic Neural networks and learning techniques but also a Fuzzy logic library and some simple machine learning algorithms like Q-learning. Furthermore the framework also contains a Vision library. The framework is also open source and is easy to use. Some very good examples can be found at codeproject.com [url="http://www.aforgenet.com/framework/"]http://www.aforgenet.com/framework/[/url]
  5. Okay, thanks Dave Mark. Hmm.. Automated planners is a whole new category of algorithm I didn’t knew about, so I think I need to make some research on the topic before I continue my work on the solution. [b]Info for other users at Gamedev:[/b] The idea of planners [url="http://en.wikipedia.org/wiki/Automated_planning"]http://en.wikipedia.org/wiki/Automated_planning[/url] STRIPS [url="http://en.wikipedia.org/wiki/STRIPS"]http://en.wikipedia.org/wiki/STRIPS[/url] GOAP ( Simplfied version of STRIPS) [url="http://web.media.mit.edu/~jorkin/goap.html"]http://web.media.mit.edu/~jorkin/goap.html[/url] Ironic.. here is an game engine named “[b]X-ray[/b]” which uses GOAP of NPC’s [img]http://public.gamedev.net//public/style_emoticons/default/smile.png[/img] [url="http://en.wikipedia.org/wiki/X-Ray_Engine"]http://en.wikipedia.org/wiki/X-Ray_Engine[/url]
  6. Thanks Jeffery, It is a great idea. By using the A* as an action planner instead of has a pathfinder, I may be able to solve the problem. It opens a whole new set of options. I could do things like make some movement cost more than others. And as you suggest, it can use my preprogrammed sequences. I think you are right (again [img]http://public.gamedev.net//public/style_emoticons/default/smile.png[/img] ) - I need to make the A* time slice based. I will start working with your suggestion right away. It will not be easy, but hopefully it will work. It would be a lot cleaner solution than with ANN. Once again thanks.
  7. Thanks for your answer Jeffery, I agree with your point about limited obstacles and I think I understand your point. The way the system currently works is by a big handmade table with current position as input. The feedback from the table is something like your description: 1. Move "arm 1" to 75 cm in Z 2. Move body to 600 cm 3. If "arms are free" "turn body toward goal" else if "possible move arm 2 up". 4. etc. I use the hit boxes show in the image to determine if a movement is possible or will result in collision. The problem with that solution is that it is not very dynamic. The "stand" can be moved and the table can be moved by the user. I have also made a system where smaller easy recognizable sub states are recognized: 1. If "Big movement" Move to body to 600 cm 2. If "arms need to switch places" do sequence 132 3. If "Goal position along tableside" do sequence 97 4. etc. It made the task a little easier but still with the same problem, it is not dynamic. In a new situation where the "stand" is put in a new position and the table is turn, I have to make a lot of new add-ons to the table without affecting the already working part of the table. Was it something like above you had in mind? Jeffery you are right about the application of the robot. [img]http://public.gamedev.net//public/style_emoticons/default/smile.png[/img] I am very open to all ideas and I am grateful for all suggestions. Edit: Sorry, I was asked to remove the picture from the forum by my boss.
  8. Thanks for your suggestions. The problem is the number of axles when it comes to controlling the motors. I have created 4 screenshot from the simulation so you are able to see have the robot works. I am very grateful for your suggestions. Edit: Sorry, I was asked to remove the pictures from the forum by my boss.
  9. Thank you Jefferytitan and slicer4ever, Well, the robot is hanging from the ceiling and moves along a rail. The two arms are operating down on floor level and shall do things like go under a table without touching it. The question is how to for example move the arms from above the table to under the table without colliding with the table and without the two arms colliding. The arms are never allowed to touch anything in the room. 1 motor is used for moving the robot along the rails and 1 motor is used to turn the main body of the robot. 7 motors in one arm and 6 motors in the other arm. So I have two problems. 1. Is the pathfinding. 2. How to follow the path with the use of the motors in the arms and the main body. Thanks again, I will look in to D* and RVO/VO and see if it is an solution I can use.
  10. Okay. Thanks for letting me post IADaveMark. If I find a solution for my problem I am hope it also will benefit game develops. The 3D simulation works as a sort of a game. [img]http://public.gamedev.net//public/style_emoticons/default/biggrin.png[/img] By the way, I found a neural network learning algorithms which I may be able to use. It is called Simulated Annealing. As with Genetic algorithms it is able to solve the Traveling Salesman problem. In that problem there are a group of cities and the network decides which is the best route to take. As I understand it Simulated Annealing just do it fast. [url="http://www.youtube.com/watch?v=kF2heJbvYcU"]http://www.youtube.com/watch?v=kF2heJbvYcU[/url] I am not sure it is the best way to go so I am open for all ideas.
  11. Yea.. it's not a game. [img]http://public.gamedev.net//public/style_emoticons/default/smile.png[/img] But I think the problem in many ways is game related. I think I need some kind of neural network so I thought AI was the right area to put the post.
  12. Hello, [b]My project:[/b] I have a 2000 pounds real life robot with two arms. The arms are controlled by 15 motors and gets position feedback from 15 potentiometers. The robot is placed in a room with stationary objects which positions are known. The motors are only able to move with full speed. [b]Task:[/b] Get the two arms of the robot from their current point to a selected point. [b]My current solution:[/b] Currently the robot is moving by a big handmade table, but the problem is that if an object is moved the robot is going to collide. And it is almost impossible to imagine all possible situations. [b]Environment:[/b] I have created a 3D simulation for the robot which I am planning to use when training the robot. The 3D simulation can detect collision with objects and between the two arms. [b]My question:[/b] First I looked at some path finding algorithms like A* but with 15 motor in a 3D environment it is not very useful. Then I begin to look at neural network and Q-learning but I am not sure it is the right way to go. I think I need some kind of reinforcement learning. My input and output could be something like: Input: 15 x Current motor pos 15 x Goto pos Obejects in room pos or the 3D simulation Output: 15 x (maybe 5 states) sub positions for the motors I am currently trying to figure out which network and learning method to use if I go with neural network and reinforcement learning. Is it doable? I hope someone can help me and maybe also point to an example. Thank you in advance. With best regards, Peter