one more quick question, as I didn't find much info in the docs on orthogonal regions: would I be correct in intuiting that they always update their states in the order declared in the region's template?
Well, @andrew-gresyk . Congratulations for the new version 1.9.0 of HFSM2!! It's been a while since I've worked with any project involving a finite state machine. But you can believe me: I'm looking forward to it. It is my desire to use HFSM2 in the context of a robotic platform I am working with. :-)
1 reply
Hi @andrew-gresyk, I've gotten a bit further into integrating HFSM2 into our robotics project; really nice to work with. There is one slightly unexpected behaviour that I encountered and I just wanted to check with you if there is a way of avoiding it. I modeled my overall HFSM into a 2-level state+substate, using Composite regions (with head states) as the states, and just states as the substates. There is some logic that i want to call every for every update of the region, so I overrode the head state's update() method and performed that logic there. This logic contains error checks, like loss of connectivity, timeouts expiring, etc, and is meant to pre-emptively exit the region upon detection of these issues. While this works, I noticed that when one of these errors occurred and I called changeTo<SomeOtherRegion> from the head's update method, the currently active substate update method got called immediately afterwards, despite the request to exit the region. Is this expected, and is there any way to "force" deactivation of a region immediately to avoid the region's substates getting updated? Perhaps there is a better way to achieve this? I'm trying to avoid putting the same error-checking code in every substate, since this logic is common to all of them.
I've tried a lot of alternative methods, like composing the "State" as an orthogonal region containing the "pre-update check" state and the original Composite region containing the substates (same behaviour as just doing the check/transition in the Composite's head state), as well as trying to use an Injection's preUpdate() method to call a lambda with a reference to the FSM that evokes the transition (the inner state's update methods still get called here as well).
Apologies if this just sounds like word soup to you :sweat_smile:
So far my only working approach is to create a map of StateID->PreUpdate std::functions and evoke "state_preupdate_functions fsm.stateId() ;" before each fsm.update() in the main loop.
But I was hoping to find a way to embed the logic into the FSM transition-related code self-contained
Hey @spiderkeys_gitlab
Thank you!
What you can do - is use event handling for handling errors.
There's an example in https://github.com/andrew-gresyk/HFSM2/blob/master/examples/advanced_event_handling/main.cpp for that.
You can have your FSM react() to CheckForErrors event before the recursive update()
I inform you that today, after so long, I have returned to HFSM2 (and the deterministic world of finite state machines) :-).
I have started by recalling a little bit my previous experience with HFSM2.
This time I want to apply it to a wheel-based mobile robotic platform. Let's see if I succeed!
At the moment what I miss the most is the documentation, which is very incomplete. Especially what has to do with the advanced features of HFSM2. :-/
Hey @DJuego
It easy to underestimate how much time it takes to write good docs.
I can definitely use some help with writing documentation, since it always ends up being the activity with the lowest payout for me :-(
For now, I'm focusing my efforts on the next library that implements a generic planner, which takes up most of my free time.
Meanwhile, if you describe your goal - I'll do my best to assist.
Also, if you have a specific use case, I can make an example to demo how that could be done, and include it in the library distribution.
It easy to underestimate how much time it takes to write good docs.
I can definitely use some help with writing documentation, since it always ends up being the activity with the lowest payout for me :-(
For now, I'm focusing my efforts on the next library that implements a generic planner, which takes up most of my free time.
Meanwhile, if you describe your goal - I'll do my best to assist.
Also, if you have a specific use case, I can make an example to demo how that could be done, and include it in the library distribution.
Hello everybody, Hello @andrew-gresyk
Thank You for great library.
I use Your library with success for embedded programming on ESP32 (xtensa core) with esp-idf.
I use the actor reference as the context, as suggested here in the chat.
Till now I have managed to overcome all problems I encountered.
But now I want to define actor as a template class and I'm struggling with this.
Thank You for great library.
I use Your library with success for embedded programming on ESP32 (xtensa core) with esp-idf.
I use the actor reference as the context, as suggested here in the chat.
Till now I have managed to overcome all problems I encountered.
But now I want to define actor as a template class and I'm struggling with this.
Here is the code
#define HFSM2_ENABLE_PLANS
#include <hfsm2/machine.hpp>
#include <doctest/doctest.h>
template<typename E>
class CtrlBTempl {
public:
CtrlBTempl():
_machine(*this){};
E f(){return E{};};
private:
//states
struct RegionDisabledState;
struct RegionEnabledState;
struct ReadyState;
struct SuspendedState;
using Config = typename hfsm2::Config::ContextT<CtrlBTempl<E>&>;
using Machine = typename hfsm2::MachineT<Config>;
using HFSM = typename Machine::template PeerRoot< RegionDisabledState,
typename Machine::template Composite<RegionEnabledState,
ReadyState,
SuspendedState
>
>;
struct RegionDisabledState : HFSM::State {
using HFSM::State::react;
// error: ‘Control’ has not been declared
// void enter(Control& control){control._().f(); };
// void react(const E&, FullControl& control) {control.changeTo<RegionEnabledState>();}
};
struct RegionEnabledState : HFSM::State {
using HFSM::State::react;
};
struct ReadyState : HFSM::State {
using HFSM::State::react;
};
struct SuspendedState : HFSM::State {
using HFSM::State::react;
};
typename HFSM::Instance _machine;
};
TEST_CASE("template_context_test")
{
CtrlBTempl<int> cbt;
}Adding using typename HFSM::State::PlanControl; to the state declarations fixes the problem.
Also, keep in mind that both enter() and exit() take PlanControl, not Control.
Here's the complete list:
void entryGuard (GuardControl&)
void exitGuard (GuardControl&)
void enter (PlanControl&)
void reenter (PlanControl&)
void exit (PlanControl&)
void update (FullControl&)
void react (const TEvent&, FullControl&)
void planSucceeded(FullControl&)
void planFailed (FullControl&)
Hello,
first of all, thanks for this great statemachine implementation. I am currently using it on embedded device too.
I was just wondering: Is there any function, that is called on every statechange inside a given region? First I thought the exit method of the regions-super-state would be called in every statechange. But it seems that this is not the case.
Is there anythin like this, I can use for this purpose? Our goal is to reset a timer on every transition, so timings from one state wont affect other states. I am aware I can achieve what I want by inserting this call all of the states inside my region. But thats a lot of boilerplate.
first of all, thanks for this great statemachine implementation. I am currently using it on embedded device too.
I was just wondering: Is there any function, that is called on every statechange inside a given region? First I thought the exit method of the regions-super-state would be called in every statechange. But it seems that this is not the case.
Is there anythin like this, I can use for this purpose? Our goal is to reset a timer on every transition, so timings from one state wont affect other states. I am aware I can achieve what I want by inserting this call all of the states inside my region. But thats a lot of boilerplate.
Here's an example: https://github.com/andrew-gresyk/HFSM2/blob/master/test/test_injections.cpp
Hi,
I experienced non deterministic application behavior using HFSM2 on ESP32. I think it was related to 2 cores of ESP32.
This manifested itself, for example, in the fact that there was no transition from one state to another even if there was no other possibility. e.g. the machine reacted to the event, but despite the react function was executed, (what could be seen in logs), and even if in react function changeTo transition was included, there was no transition to other state.
I suspect that preemption took place at that time and when there was a return to the place of execution of the program in which the transition should take place, the program did not work properly.
I pinned all my machines to only one core of CPU and now I can't reproduce this issue anymore. Previosuly it was quaite easy to reproduce.
Does anyone have experience in using HFSM2 in mulitcore embeded system? I would like to be sure that I realy solved my issue so any hint how to prove this is welcome.
@andrew-gresyk You probaly use your framework in multicore enviroment on desktop
I experienced non deterministic application behavior using HFSM2 on ESP32. I think it was related to 2 cores of ESP32.
This manifested itself, for example, in the fact that there was no transition from one state to another even if there was no other possibility. e.g. the machine reacted to the event, but despite the react function was executed, (what could be seen in logs), and even if in react function changeTo transition was included, there was no transition to other state.
I suspect that preemption took place at that time and when there was a return to the place of execution of the program in which the transition should take place, the program did not work properly.
I pinned all my machines to only one core of CPU and now I can't reproduce this issue anymore. Previosuly it was quaite easy to reproduce.
Does anyone have experience in using HFSM2 in mulitcore embeded system? I would like to be sure that I realy solved my issue so any hint how to prove this is welcome.
@andrew-gresyk You probaly use your framework in multicore enviroment on desktop
If are using HFSM2 in a multithreaded environment, it should probably be protected by a critical section, or some other thread / process synchronization primitive.
@andrew-gresyk!, I didn't know you were Ukrainian! Now I just realized that the colors of the *FSM2 logo are those of the ukrainian flag.