I am working on a reinforcement learning environment for a robot using gz-sim8, and I am trying to implement the reset functionality. I would like to call my resets with a ResetSim() function, as seen here. I am currently trying to call a reset similar to the GUI's reset button with the gz.msgs.WorldControl service call to /world/<world_name>/control. Seeing the reset button work in my gui when I run gz sim empty.world makes me believe I can reset the simulation with service calls, as seen with my code:
void TrainSimulator::ResetSim(){
gz::transport::Node node;
// Create a message to reset the world
gz::msgs::WorldControl req;
req.mutable_reset()->set_all(true); // Reset everything
gz::msgs::Boolean rep;
std::string service = "/world/empty/control";
bool result = false;
unsigned int timeout = 5000; // Timeout in milliseconds
// Request reset through the transport node
bool success = node.Request(service, req, timeout, rep, result);
if (!success || !result || !rep.data()) {
std::cerr << "Failed to reset the world using transport service." << std::endl;
} else {
std::cout << "World reset successfully through transport service." << std::endl;
}
}
Unfortunately, I get a warning in terminal that prints 3 times:
[Wrn] [SystemManager.cc:185] In-memory without ISystemReset detected: []
Systems created without plugins that do not implement Reset will not be reloaded. Reset may not work correctly
Then I get a Segmentation fault (Core dumped). I have 3 plugins loaded by my gz::sim::Server object, and I wonder if I have to remember to implement an ISystemReset for each...would It just be like the ISystemConfigure call I created?
Alternatives
I have tried resetting my simulation by resetting the robot's pose by setting gz::sim::components::WorldPoseCmd to an initial pose, setting the JointPositionReset, JointVelocityReset, JointForceCmd components to zero, and setting the AngularAcceleration, LinearAcceleration components of the model to zero. Calling this implementation of ResetSim() would make my sdf model move in erratic ways, so I tried setting AngularVelocityCmd and LinearVelocityCmd to zero to keep the robot from moving randomly, but the robot would just freeze in mid air while the simulation would run.
I am wondering what I would need to do to be able to reset my robot properly, similar to how pybullet has its restoreState() function where you can reset the world to a state that you save.
