can't pickle bark.core.models.behavior.BehaviorModel objects

[flxebert]

I'm trying to reproduce the developing behavior model tutorial in Ubuntu 18.04 and conda install. In both the pip install and the source installation execution fails with the error can't pickle bark.core.models.behavior.BehaviorModel objects.

I couldn't find any documentation or issue related to that topic. Do I miss something? How can I fix this?

Here is the full output output.txt

[juloberno]

The tutorial code is similar to the test cases implemented in bark/models/tests/py_behavior_model_test.py

You can have a look there and see if your implementation follows the test implementation. If errors persist, please add your code here and describe your build and execution pipeline.

[flxebert]

The problem occurs if the derived behavior model is returned from the _controlled_behaviormodel() in the CustomLaneCorridorConfig class. The implementation of DerivedBehaviorModel follows the ones in bark/models/tests/py_behavior_model_test.py. I followed the installation tutorial, working in a conda environment with python 3.7 and installed bark via pip install bark-simulator.

Ubuntu 18.04 python 3.7.11 bark 1.4.9

Running the following code (which comes from the tutorial) leads to the pickle error

import numpy as np
from bark.core.world.opendrive import *
from bark.core.world.goal_definition import *

from bark.core.models.behavior import BehaviorModel
from bark.core.models.behavior import BehaviorMobilRuleBased

from bark.runtime.runtime import Runtime
from bark.examples.paths import Data
from bark.runtime.commons.parameters import ParameterServer
from bark.runtime.viewer.matplotlib_viewer import MPViewer
from bark.runtime.scenario.scenario_generation.config_with_ease import \
  LaneCorridorConfig, ConfigWithEase

params = ParameterServer()

class DerivedBehaviorModel(BehaviorModel):
  def __init__(self, params=None):
    super(DerivedBehaviorModel, self).__init__(params)

  def Plan(self, step_time, observed_world):
    super(DerivedBehaviorModel, self).ActionToBehavior(np.array([2., 1.], dtype=np.float32))
    trajectory = np.array([[0., 0., 0., 0., 0.], [0., 0., 0., 0., 0.]], dtype=np.float32)
    super(DerivedBehaviorModel, self).SetLastTrajectory(trajectory)
    return trajectory

  def Clone(self):
    return self

class CustomLaneCorridorConfig(LaneCorridorConfig):
  def __init__(self, params=None, **kwargs):
    super(CustomLaneCorridorConfig, self).__init__(params, **kwargs)

  def goal(self, world):
    road_corr = world.map.GetRoadCorridor(self._road_ids, XodrDrivingDirection.forward)
    lane_corr = self._road_corridor.lane_corridors[0]
    return GoalDefinitionPolygon(lane_corr.polygon)

  def controlled_behavior_model(self, world):
    return DerivedBehaviorModel(self._params)

# configure both lanes
left_lane = CustomLaneCorridorConfig(params=params,
                                     lane_corridor_id=0,
                                     road_ids=[0, 1],
                                     behavior_model=BehaviorMobilRuleBased(params),
                                     s_min=0.,
                                     s_max=50.)

# this lane has controlled_ids; ego vehicle is placed on this lane
right_lane = CustomLaneCorridorConfig(params=params,
                                      lane_corridor_id=1,
                                      road_ids=[0, 1],
                                      controlled_ids=True,
                                      behavior_model=BehaviorMobilRuleBased(params),
                                      s_min=0.,
                                      s_max=20.)

# finally, generate 3 scenarios on the merging map
scenarios = \
  ConfigWithEase(
    num_scenarios=3,
    map_file_name=Data.xodr_data("DR_DEU_Merging_MT_v01_shifted"),
    random_seed=0,
    params=params,
    lane_corridor_configs=[left_lane, right_lane])

viewer = MPViewer(params=params,
                  x_range=[-35, 35],
                  y_range=[-35, 35],
                  follow_agent_id=True)
env = Runtime(step_time=0.2,
              viewer=viewer,
              scenario_generator=scenarios,
              render=True)

# run 3 scenarios
for _ in range(0, 3):
  env.reset()
  # step scenario 90 time-steps
  for step in range(0, 90):
    env.step()