I am trying to build a simple multibody plant system in Drake using the basic DrakeVisualizer. However, for my use case, I also want to be able to automatically track the derivatives through the physics simulation, so am using the AutoDiffXd version of system:
timestep = 1e-3
builder = DiagramBuilder_[AutoDiffXd]()
plant = MultibodyPlant(timestep)
scene_graph = SceneGraph_[AutoDiffXd]()
brick_file = FindResourceOrThrow("drake/examples/manipulation_station/models/061_foam_brick.sdf")
parser = Parser(plant)
brick = parser.AddModelFromFile(brick_file, model_name="brick")
plant.Finalize()
plant_ad = plant.ToAutoDiffXd()
plant_ad.RegisterAsSourceForSceneGraph(scene_graph)
scene_graph.AddRenderer("renderer", MakeRenderEngineVtk(RenderEngineVtkParams()))
DrakeVisualizer.AddToBuilder(builder, scene_graph)
builder.AddSystem(plant_ad)
builder.AddSystem(scene_graph)
builder.Connect(plant_ad.get_geometry_poses_output_port(), scene_graph.get_source_pose_port(plant_ad.get_source_id()))
builder.Connect(scene_graph.get_query_output_port(), plant_ad.get_geometry_query_input_port())
diagram = builder.Build()
context = diagram.CreateDefaultContext()
simulator = Simulator_[AutoDiffXd](diagram, context)
simulator.AdvanceTo(2.0)
However, when I run this, I get the following error:
File "/home/craig/Repos/drake-exps/autoDiffExperiment.py", line 102, in auto_phys
DrakeVisualizer.AddToBuilder(builder, scene_graph)
TypeError: AddToBuilder(): incompatible function arguments. The following argument types are supported:
1. (builder: pydrake.systems.framework.DiagramBuilder_[float], scene_graph: drake::geometry::SceneGraph<double>, lcm: pydrake.lcm.DrakeLcmInterface = None, params: pydrake.geometry.DrakeVisualizerParams = <pydrake.geometry.DrakeVisualizerParams object at 0x7ff6274e14b0>) -> pydrake.geometry.DrakeVisualizer
2. (builder: pydrake.systems.framework.DiagramBuilder_[float], query_object_port: pydrake.systems.framework.OutputPort_[float], lcm: pydrake.lcm.DrakeLcmInterface = None, params: pydrake.geometry.DrakeVisualizerParams = <pydrake.geometry.DrakeVisualizerParams object at 0x7ff627736730>) -> pydrake.geometry.DrakeVisualizer
Invoked with: <pydrake.systems.framework.DiagramBuilder_[AutoDiffXd] object at 0x7ff65654f8f0>, <pydrake.geometry.SceneGraph_[AutoDiffXd] object at 0x7ff656562130>
From this error, it appears the DrakeVisualizer class only accepts systems which use float scalars exlusively. So I am stuck --- either I can go back to floats (but lose the autodiff differentiable simulation functionality I was after in the first place), or continue to use autodiffxd systems (but be completely unable to visualize what is going on in my simulation).
Is there a way to get both that I am missing?
Sorry for the pain and inconvenience. Your description and assessment are all spot on. Most of the visualization mechanisms are float only and, in its current state, attempts to visualizing an AutoDiff diagram will fail.
You have a couple of options (neither of which is appealing):
Go with one of the outcomes you've described above (no vis or no derivatives).
Put in a Drake feature request to be able to attach a visualizer to an AutoDiff diagram.
I can come up with some hacky workarounds (that aren't immediately clear would even work). So, if you're desperate for derivatives and visualization, they could be explored. But, ultimately, the feature request and a formal Drake solution would be the best long-term resolution.
=====================================
Big update. As of #14569, the DrakeVisualizer class is now templated on the scalar type (item 2 in the list above). That has two implications:
You can build an AutoDiffXd-valued diagram with a visualizer in it (as in your example), or
You can create a double-valued diagram and scalar convert it (i.e., diagram.ToAutoDiffXd() into an AutoDiffXd-valued diagram.
Related
So I have created a huge screen that essentially just shows the robot status for every robot in this factory (individually)… At the very end of the project, they decided they want one object on the screen that blinks if any of the 300 robots fault. I am trying to think of a way to make this work. Maybe a global script of some kind? Problem is, I do not do much scripting in Cimplicity, so any help is appreciated.
All the points that are currently used on this screen (to indicate a fault) have very similar names… as in, the beginning is the same… so I was thinking of a script that could maybe recognize if a bit is high based on PART of it's string name characteristic. The end will change a little each time, but I am sure there is a way to only look for part of a string and negate the rest. If the end has to be hard coded, that's fine.
You can use a Python script in Cimplicity.
I will not go into detail on the use of python in Cimplicity, which is well described in the documentation indicated above.
Here's an example of what can be done... note that I don't have a way to test it and, of course, this will work if the name of your robots in the declaration follows the format Robot_1, Robot_2, Robot_3 ... Robot_10 ... Robot_300 and it also depends on the Name and the Type of the fault variable... as you didn't define it, I imagine it can be an integer, with ZERO indicating no error. But if you use something other than that, you can easily change it.
import cimplicity
(...)
OneRobotWithFault = False
# Here you get the values and check for fault
for i in range(0, 300):
pointName = f'MyFactory.Robot_{i}.FaultCode'
robotFaultCode = cimplicity.point_get(pointName)
if robotFaultCode > 0:
OneRobotWithFault = True
break
# Set the status to the variable "WeHaveRobotWithFault"
cimplicity.point_set("WeHaveRobotWithFault", OneRobotWithFault)
P.S. This Question has been edited to answer questions made by #Felipe
I have an Agent-Based model simulation for churn behavior modeling. On each iteration(based on time--month) each user reconsiders her choice of operator(our or other) based on model metrics (Cost/SocialNetwork/...). In runtime even when I change parameters to affect Agents' decision, no one changes his/her operator. here is my state chart image on the below:
I should note that internal transition of (our user) has below details:
the first two lines are something for display. Advocate() refers to the action of sending messages which affects social influence.
But Switch() is where decision happens based on new parameters' value. In short, d defines a normalized range between -1 and 1 : signum(d) predicts which provider is the preferred one and abs(d) shows how preferred the selected provider will be.
//Definition for Switch()
double d = (this.Social_impact()/20)+this.Monthly_Charge_Impact();
if (d>0)
SwitchToUs();
else
SwitchToOther();
the two SwitchToUs and SwitchToOther functions simply change the operator (as if creating arrows between OUR_USER and OTHER_USER states)
I work with pyfmi in Jupyter notebooks to run simulations and I like to work interactively and evaluate incremental changes in parameters etc. Long time ago I found it necessary to introduce a dictionary that work as a "container" for parameter and initial values. Now I wonder if here is a way to get rid of this "container" that after all is partly a parallel structure to "model"?
A typical workflow look like this:
create a diagram where results from different simulations below should be shown
model = load_fmu(fmu_model)
parDict['model.x_0'] = 1
parDict['model.a'] = 2
for key in parDict.keys(): model.set(key,parDict[key])
sim_res = model.simulate(10)
plot results...
model = load_fmu(fmu_model)
parDict['model.x_0'] = 3
for key in parDict.keys(): model.set(key,parDict[key])
sim_res = model.simulate(10)
plot results...
There is a function model.reset() that brings the state back to default values at compilation without loading again, but you need to do more than the following
model.reset()
parDict['model.x_0'] = 3
for key in parDict.keys(): model.set(key,parDict[key])
sim_res = model.simulate(10)
plot results...
So, this does NOT work...
and after all parameters and initial values needs to be brought back and we still need parDict, but we may avoid the load-command though.
I'm having a problem when I use the "Split model" option. What I want to do is basically hide these 10 water volumes:.
I select the tanks then I click on button for splitting with these options:
Final result is just what I want:
When I check the entire model to verify if everything is ok, these errors come out:
I've tried several things such as modifying the text part of the splitted model with no positive results, here's the original NOT modified
Can you please explain to me what kind of error it is? How can I resolve it? Thank you.
Edit: I'm using TIL library
Edit after Markus' answer: in the split model is it necessary to declare the type of liquid and change the portArray definition. I copied these lines of code and everything worked!
parameter TILMedia.LiquidTypes.BaseLiquid liquidType = sim.liquidType1
"Liquid type" annotation (Dialog(tab="SIM",group="SIM"),choices(
choice=sim.liquidType1 "Liquid 1 as defined in SIM",
choice=sim.liquidType2 "Liquid 2 as defined in SIM",
choice=sim.liquidType3 "Liquid 3 as defined in SIM"));
replaceable package MediaConfiguration =
TIL.Utilities.MediaConfiguration
constrainedby TIL.Utilities.Internals.PartialMediaConfiguration
"Media and State Type Configuration" annotation (choicesAllMatching, Dialog(
tab="SIM", group="Media Configuration"));
protected
outer TIL.SystemInformationManager sim "System information manager";
and
public
TIL.Connectors.LiquidPort portArray(
final liquidType=liquidType) ;
TIL.Connectors.LiquidPort portArray1(
final liquidType=liquidType) ;
The issue seems to result from the vectorization of the connectors, that seems to get lost when using "split model". A bit difficult without the actual model, but:
Have you tried to modify the last two connect statements in str3000 to:
connect(portArray, colume.portArray[1])
connect(portArray1, colume.portArray[2])
Additionally on the top level of the model, it seems you have connections to vectors of str3000.portArray. Try to remove them as they seem to be wrong, as you have two non-vector ports.
There should be something like connect(str3000.portArray[1], ...) and connect(str3000.portArray1[2], ...), which should likely be changed to connect(str3000.portArray, ...) and connect(str3000.portArray1, ...).
I am able to create and optimize neuralODEs in julia(1.3 and 1.2) using Flux.jl and DiffEqFlux.jl but it fails under a crucial important general case.
what works:
I can train the Neural net parameters if it is built out of the
provided Flux.jl layers like Dense().
I can include an arbitrary function as a layer in the network chain, e.g. x -> x.*x
What fails:
However if the arbitrary function has parameters I want to train then Flux. Train will not adjust these parameters causing it to fail.
I have tried making these added parameters Tracked and included in the list of parameters given to the training system but it ignores them and they remain unvaried.
The documentation says very cryptically that one can use Flux.#functor on a layer to make sure it's parameters get tracked. However functor was not included in Flux till version 0.10.0 and the only version of Flux compatible with NeuralODEs in DiffEqFlux is 0.9.0
So here's an toy example of a 2 layer neural net I want to use
p = param([1.0])
dudt = chain( x -> p[1]*x.*x, Dense(2,2) )
ps = Flux.params(dudt)
then I use the flux train on this. when I do this the parameter p is not varied, but the parameters in the Dense layer are.
I have tried explicitly including like this
ps = Flux.Params([p,dudt])
but that has the same result and the same problem
I think what I need to do is build a struct with an associted function that implements the
x->p[1]*x*x
then call #functor on this. That struct can then be used in the chain.
But as I noted the version of Flux with #functor is not compatible with DiffEqFlux of any version.
So I need a way to make flux pay attention to my custom parameters, not just the ones in Dense()
How???
I think I get what your question is, but please clarify if I am answering the wrong question here. The issue is that the p is only grabbing from a global reference and thus not differentiated during adjoints. A much better way to handle this in 2020 is to use FastChain. The FastChan interface lets you define layer functions and their parameter dependencies, so this is a nice way to make your neural network incorporate arbitrary functions with parameters. Here's what that looks like:
using DifferentialEquations
using Flux, Zygote
using DiffEqFlux
x = Float32[2.; 0.]
p = Float32[2.0]
tspan = (0.0f0,1.0f0)
mylayer(x,p) = p[1]*x
DiffEqFlux.paramlength(::typeof(mylayer)) = 1
DiffEqFlux.initial_params(::typeof(mylayer)) = rand(Float32,1)
dudt = FastChain(FastDense(2,50,tanh),FastDense(50,2),mylayer)
p = DiffEqFlux.initial_params(dudt)
function f(u,p,t)
dudt(u,p)
end
ex_neural_ode(x,p) = solve(ODEProblem(f,x,tspan,p),Tsit5())
solve(ODEProblem(f,x,tspan,p),Tsit5())
du0,dp = Zygote.gradient((x,p)->sum(ex_neural_ode(x,p)),x,p)
where the last value of p is the one parameter for p in mylayer. Or you can directly use Flux:
using DifferentialEquations
using Flux, Zygote
using DiffEqFlux
x = Float32[2.; 0.]
p2 = Float32[2.0]
tspan = (0.0f0,1.0f0)
dudt = Chain(Dense(2,50,tanh),Dense(50,2))
p,re = Flux.destructure(dudt)
function f(u,p,t)
re(p[1:end-1])(u) |> x-> p[end]*x
end
ex_neural_ode() = solve(ODEProblem(f,x,tspan,[p;p2]),Tsit5())
grads = Zygote.gradient(()->sum(ex_neural_ode()),Flux.params(x,p,p2))
grads[x]
grads[p]
grads[p2]