go_away

Author Topic: Software-only prototyping - feasible?  (Read 1159 times)

0 Members and 1 Guest are viewing this topic.

Offline mviljamaaTopic starter

  • Jr. Member
  • **
  • Posts: 11
  • Helpful? 0
Software-only prototyping - feasible?
« on: June 25, 2013, 04:33:12 PM »
I was wondering whether anyone understands and could describe the limitations or as well the possibilities of robotics simulations to a beginner. How useful is simulation and "how far" can one reason (particularly about mechanics and control) with just simulations (i.e. no actual hardware involved at all)? Any examples of software or toolsets (involving many different software) that enable very high level of flexibility for a robot's design phase? What are the drawbacks of simulation (is it slow? inaccurate?)?
« Last Edit: June 25, 2013, 05:17:17 PM by mviljamaa »

Offline newInRobotics

  • Supreme Robot
  • *****
  • Posts: 1,015
  • Helpful? 48
  • N.I.R.
Re: Software-only prototyping - feasible?
« Reply #1 on: June 26, 2013, 12:50:15 AM »
Simulation is neither slow or inaccurate, it is much faster and cheaper than physical build, hence people do them to save time, money and to detect mistakes/errors while still in design phase. Separate simulations are done for separate disciplines involved in robot building (simulations of signal acquisition and processing, simulations of closed/open loop response and control, simulations of parts' rigidity/flexibility, simulations of manufacturing processes, etc.), as opposed to one-off all-in-one simulation covering every aspect of the robot.

For control simulations You probably want to get something like MATLAB, as for mechanics simulation - something like Inventor, SolidWorks or CATIA.

Main drawback of simulations is that they are perfectly precise and never-changing. This fact forces engineers to make decisions based on real-world experience, as real-world is always-changing and no 2 grains of sand (or anything else) are identical.

Also, another drawback is that CAD related simulations normally require some hard-core computers to run (at JLR, for manufacturing simulations we use 3.1GHz 12 CPU, 50GB of RAM, NVIDIA Quadro 5000 beasts).
"Computer games don't affect kids; I mean if Pac-Man affected us as kids, we'd all be running around in darkened rooms, munching magic pills and listening to repetitive electronic music." - Kristian W

Offline mviljamaaTopic starter

  • Jr. Member
  • **
  • Posts: 11
  • Helpful? 0
Re: Software-only prototyping - feasible?
« Reply #2 on: June 26, 2013, 05:25:59 AM »
Simulation is neither slow or inaccurate, it is much faster and cheaper than physical build, hence people do them to save time, money and to detect mistakes/errors while still in design phase. Separate simulations are done for separate disciplines involved in robot building (simulations of signal acquisition and processing, simulations of closed/open loop response and control, simulations of parts' rigidity/flexibility, simulations of manufacturing processes, etc.), as opposed to one-off all-in-one simulation covering every aspect of the robot.

For control simulations You probably want to get something like MATLAB, as for mechanics simulation - something like Inventor, SolidWorks or CATIA.

Main drawback of simulations is that they are perfectly precise and never-changing.

Thank you for your answer.

So this means that software-only prototyping for robot systems and their parts is in general very feasible? What I've been mostly concerned about has been the mechanics simulation, because as you say the virtual environment and the real environment can differ and usually differ a lot (particularly outdoors).

It's interesting how the process continues from the simulations to real-world tests though and it's something that I would like to know more about. I might be asking a common question, but what are some resources e.g. books that might provide insight to the software-based prototyping and how it through iterations leads to an actual physical robot or at least the specifications for a physical robot?
« Last Edit: June 26, 2013, 05:52:07 AM by mviljamaa »

Offline newInRobotics

  • Supreme Robot
  • *****
  • Posts: 1,015
  • Helpful? 48
  • N.I.R.
Re: Software-only prototyping - feasible?
« Reply #3 on: June 26, 2013, 06:15:44 AM »
It would be a lot easier to answer Your questions if You told us what exactly You want to simulate, as simulation of mechanics is quite broad.
"Computer games don't affect kids; I mean if Pac-Man affected us as kids, we'd all be running around in darkened rooms, munching magic pills and listening to repetitive electronic music." - Kristian W

Offline mviljamaaTopic starter

  • Jr. Member
  • **
  • Posts: 11
  • Helpful? 0
Re: Software-only prototyping - feasible?
« Reply #4 on: June 26, 2013, 06:32:14 AM »
It would be a lot easier to answer Your questions if You told us what exactly You want to simulate, as simulation of mechanics is quite broad.

Well I'm looking to virtually prototype the structure/mechanics and movement for a low-speed outdoor robot (for a computer vision application) that would move in a rough terrain and I'm specifically trying to find the cheapest and simplest design for it to be functional (in the rough terrain) prior to even looking for specific parts or materials.

What I'm unsure about is the leap between the simulations and the "real deal" and how simulations can be or should be used to predict the known conditions (the rough terrain), which is why I made this thread to consult about the "general feasibility" and accuracy of simulation. I'm also curious about the software packages that lessen the need to reinvent the wheel (e.g. by including or being able to import existing parts) + I'm specifically interested in using open source software.
« Last Edit: June 26, 2013, 06:49:20 AM by mviljamaa »

Offline newInRobotics

  • Supreme Robot
  • *****
  • Posts: 1,015
  • Helpful? 48
  • N.I.R.
Re: Software-only prototyping - feasible?
« Reply #5 on: June 26, 2013, 08:29:34 AM »
What do You consider functional? What are the problems You foresee and want to check in simulation? Do You have preliminary design of a system in question to start Your simulation on? Which part of mechanical system do You want to simulate, is it damper response time, is it motor heat dissipation, is it traction with surface, is it chassis flex, etc.? I'll repeat myself - there is no software that would give You "perfect" overall design from nothing. You have to create something first, then test it, then refine it, test it again, refine it again, etc., etc.
"Computer games don't affect kids; I mean if Pac-Man affected us as kids, we'd all be running around in darkened rooms, munching magic pills and listening to repetitive electronic music." - Kristian W

Offline mviljamaaTopic starter

  • Jr. Member
  • **
  • Posts: 11
  • Helpful? 0
Re: Software-only prototyping - feasible?
« Reply #6 on: June 26, 2013, 08:45:54 AM »
What do You consider functional? What are the problems You foresee and want to check in simulation? Do You have preliminary design of a system in question to start Your simulation on? Which part of mechanical system do You want to simulate, is it damper response time, is it motor heat dissipation, is it traction with surface, is it chassis flex, etc.? I'll repeat myself - there is no software that would give You "perfect" overall design from nothing. You have to create something first, then test it, then refine it, test it again, refine it again, etc., etc.

I have nothing designed currently, but I'll be studying existing designs as soon as I have some kind of a prototyping platform (i.e. a mechanics/physics simulator) to test out ideas.

My first task for the robot would be about the structure, dimensions and method of movement (e.g. what type of wheeling, how many wheels and what kind of suspension) to assess a simple structure that can function (move, does not get stuck and can move in a relatively balanced manner) in a rough terrain. After this it would be about further parts optimization, the power source, control, materials and all the other jazz.

Thus what I'm looking for is a software toolset in which I can do assessments about how different structural designs differ and how they can be improved or optimized for my particular needs. Conceptually I'm looking for information resources that would provide guidelines as to what you need to import to the simulations in order to get the kind of simulations that correspond closely to the real environment that the physical robot will be dealing with.
« Last Edit: June 26, 2013, 08:48:23 AM by mviljamaa »

Offline newInRobotics

  • Supreme Robot
  • *****
  • Posts: 1,015
  • Helpful? 48
  • N.I.R.
Re: Software-only prototyping - feasible?
« Reply #7 on: June 27, 2013, 12:38:29 AM »
I have nothing designed currently, but I'll be studying existing designs as soon as I have some kind of a prototyping platform (i.e. a mechanics/physics simulator) to test out ideas.

My first task for the robot would be about the structure, dimensions and method of movement (e.g. what type of wheeling, how many wheels and what kind of suspension) to assess a simple structure that can function (move, does not get stuck and can move in a relatively balanced manner) in a rough terrain.
If I were You, I'd start with modeling robot using Inventor (free for students) and using its kinematics simulator to visualize how suspension moves. Also, Inventor offers stress simulation for You to see if part is going to flex under load and how much.

As for environment/physics simulator - the one that I know is Microsoft Robotics Developer Studio. I've never used it and don't really know how difficult or easy it is to use, or how realistic simulations are.
"Computer games don't affect kids; I mean if Pac-Man affected us as kids, we'd all be running around in darkened rooms, munching magic pills and listening to repetitive electronic music." - Kristian W

Offline mviljamaaTopic starter

  • Jr. Member
  • **
  • Posts: 11
  • Helpful? 0
Re: Software-only prototyping - feasible?
« Reply #8 on: June 27, 2013, 06:00:26 AM »
If I were You, I'd start with modeling robot using Inventor (free for students) and using its kinematics simulator to visualize how suspension moves. Also, Inventor offers stress simulation for You to see if part is going to flex under load and how much.

As for environment/physics simulator - the one that I know is Microsoft Robotics Developer Studio. I've never used it and don't really know how difficult or easy it is to use, or how realistic simulations are.


Thank you for the suggestions. I've considered adopting V-Rep for the simulation work as it seems quite inclusive, requiring only a separate CAD package e.g. FreeCAD for more effective (the built-in part editor is a bit minimalistic) parts design. For FEM analysis etc. there seem to be loads of free/open source programs e.g. Elmer.

If anyone else comes to this thread there would be a few questions that would still be interesting for discussion:

-What methodology, tools and information resources exist for virtually simulating robotics in close to their designated environments (e.g. in my case the rough terrain)?

Offline jwatte

  • Supreme Robot
  • *****
  • Posts: 1,343
  • Helpful? 81
Re: Software-only prototyping - feasible?
« Reply #9 on: June 27, 2013, 01:51:57 PM »
I build my own solution using game engines. I like the ODE simulation engine because I've used it for 15 years, but there are others (Bullet Physics, Havok, PhysX, etc.)
You can also use a complete game engine -- anything from Unreal Developer Kit to Unity 3D can probably do certain kinds of simulation for you.

Offline fireintheit

  • Jr. Member
  • **
  • Posts: 14
  • Helpful? 0
Re: Software-only prototyping - feasible?
« Reply #10 on: June 27, 2013, 10:22:53 PM »
A good place to start understanding simulations and how they work is by reading Statistics Book. There are also many stat online tutorials that will give you the basic mathematics behind simulations. The main reason people use simulation is to predict outcomes or detect mistakes. 

Offline jwatte

  • Supreme Robot
  • *****
  • Posts: 1,343
  • Helpful? 81
Re: Software-only prototyping - feasible?
« Reply #11 on: June 28, 2013, 11:49:06 AM »
Statistics isn't going to help simulate robots, though. You need mechanical dynamics for that. Which, in turn, needs multidimensional algebra. The game engines / physics simulation packages do some amount of work to hide the nastier pieces (inertia tensor, Jacobian constraint solvers) for you, but you still need to understand position-and-orientation-in-3D-as-matrix, etc.

Offline nrox

  • Jr. Member
  • **
  • Posts: 8
  • Helpful? 0
Re: Software-only prototyping - feasible?
« Reply #12 on: September 03, 2013, 02:07:26 PM »
Hi,
I released recently this online tool for robot modeling and simulation. If you have the time check it tout and leave your comments:
https://assemblino.com/
Best,
Nuno

Offline CJAlbertson

  • Jr. Member
  • **
  • Posts: 19
  • Helpful? 1
Re: Software-only prototyping - feasible?
« Reply #13 on: November 16, 2013, 03:20:21 AM »
Any examples of software or toolsets (involving many different software) that enable very high level of flexibility for a robot's design phase? What are the drawbacks of simulation (is it slow? inaccurate?)?


Sounds like you want to use Gazebo.   
http://gazebosim.org/wiki/Overview
 
Chris A. in So. Cal.

 


Get Your Ad Here