AirSimExe raises SIGABRT on incorrect vertices size

[layman-n-ish]

AirSimExe raises signal 6 (SIGABRT) on an assertion fail:
F1106 23:44:11.933295 31485 vertex.cpp:257] Check failed: vertices.size() >= 2 (1 vs. 2)

This signal is raised well into the execution of my code i.e. after ~3-4 hrs of training and I fail to recognise the source. Any help would be appreciated.

Please find attached the complete log and crash files.
GoD_crash.log
Diagnostics.txt

The client side ceases with:

Traceback (most recent call last):
  File "/usr/lib/python3.6/threading.py", line 916, in _bootstrap_inner
    self.run()
  File "/usr/lib/python3.6/threading.py", line 864, in run
    self._target(*self._args, **self._kwargs)
  File "/home/nishant/Desktop/GoD/baselines/baseline_racer.py", line 215, in repeat_timer_odometry_callback
    task()
  File "/home/nishant/Desktop/GoD/airsim_gym/envs/AirSim_Client.py", line 44, in odometry_callback
    drone_state = self.airsim_client_odom.getMultirotorState(vehicle_name=self.drone_name).kinematics_estimated
  File "/home/nishant/.local/lib/python3.6/site-packages/airsimneurips/client.py", line 1053, in getMultirotorState
    return MultirotorState.from_msgpack(self.client.call('getMultirotorState', vehicle_name))
  File "/home/nishant/.local/lib/python3.6/site-packages/msgpackrpc/session.py", line 41, in call
    return self.send_request(method, args).get()
  File "/home/nishant/.local/lib/python3.6/site-packages/msgpackrpc/future.py", line 43, in get
    raise self._error
msgpackrpc.error.TimeoutError: Request timed out 

Just to confirm, the timeout raised here is due to the server side being terminated, right?

[madratman]

Are you using moveOnSpline? If you are passing an empty waypoint (first argument), this can be a cause.
Can you give us a reproducible script? Or share which APIs you're using in the least?

[layman-n-ish]

I use moveOnSpline() only during takeoff. Thereafter, I use moveByVelocity().

The termination has occurred only 2-3 times now; I sense the behaviour is very erratic. Will still try to reproduce it and upload the script here.

[madratman]

Hmm, try putting an assert to whatever you are sending to moveOnSpline and make sure the waypoint makes sense. Also, see comment here #116 (comment)

[madratman]

@layman-n-ish any news? If you can give us a reproducible script, it would really help in nailing this down, or else the cause should be similar to my comment in issue 116 (linked above)

[layman-n-ish]

Sorry for the late reply, was away.

Are you sure this assertion fail is from moveOnSpline()? Because, as I said, I use moveOnSpline() only in the takeoff API - takeoff_with_moveOnSpline() (that has the add_position_constraint set as True) and I use moveByVelocityAsync() thereafter.

takeoff_with_moveOnSpline() is called on reset which is similar to the dummy_reset() function you provided earlier:

def _reset(self):
		self.airsim_client.simPause(True)
		self.airsim_client.reset()
		# time.sleep(0.5)
		self.airsim_client.enableApiControl(vehicle_name=self.drone_name)
		self.airsim_client.arm(vehicle_name=self.drone_name)
		self.airsim_client.simPause(False) 
		self.airsim_client.simResetRace()
		self.airsim_client.simStartRace(tier=self.race_tier)
		self.takeoff_with_moveOnSpline() 

On the otherhand, I'll work on the reproducible script.

[madratman]

@layman-n-ish see my last 3 comments in this thread #123 . It should fix your issue as well. Please try using the reset function I added in the gist here https://gist.github.com/madratman/e617b53ec20c5f38a7d10633ba3a42c9#file-airsim_neurips_reset_episode_crash_repro-py-L43

        self.airsim_client.reset()
        self.airsim_client.simResetRace()
        time.sleep(1.0)
        self.airsim_client.simStartRace(tier=1)

[layman-n-ish]

Thank you for the solution. I am trying it out now. Seems promising till now. Will let you know how it goes.

[layman-n-ish]

Unfortunately, I get the same error and this time, even earlier into my training. This is the suggested reset_episode() I used:

def _reset(self):
		self.airsim_client.reset()
		self.airsim_client.enableApiControl(vehicle_name=self.drone_name)
		self.airsim_client.arm(vehicle_name=self.drone_name)
		self.airsim_client.simResetRace()
		time.sleep(1.0)
		self.airsim_client.simStartRace(tier=self.race_tier)
		self.takeoff_with_moveOnSpline(takeoff_height=0.4) 

If the error message, as you hypothesised (#123 (comment)), is coming from drone_2 (when it calls moveOnSplineAsync()), then disabling drone_2 (#71 (comment)) can prove this hypothesis if the crash stops. Else, I can safely assume the crashes occur during resetting of an episode (in my case) when I call moveOnSplineAsync() in takeoff_with_moveOnSpline(). I tried disabling drone_2 and execute my code; no crashes yet. Will let you know how it proceeds.

[madratman]

#123 (comment)