07 Frequent Questions Archive

To be updated!

Assignment 1

Programming Tasks

• Coding fields:
  • For task 1 (parking task), please only code in the three parts with "FILL ME".
  • For task 2 (crossing task), you can add codes outside the "FILL ME" field or even modify some codes in the template if necessary.
• Error: No such file or directory: 'sas_plan':
  • make sure to call the gen.generateDomainPDDL() function at the end of generateDomainPDDLFile() function.
  • otherwise, this means the pddl solver cannot find a solution for the problem description you generated, please check your codes and logic.
• In task 2, will the agent loop like the cars?
  • No, only the cars loop.
• How to set the agent speed to (-3, -1) in task 2?
  • You need to manually change the agent_speed_range when making the evnrionment (in test function):

  env = gym.make('GridDriving-v0', lanes=LANES, width=WIDTH, random_seed=RANDOM_SEED, agent_speed_range=(-3, -1))
  

• Why doesn't my newly added actions work in task 2?
  • You need to add your action names in both simulateSolution and generatePlan functions as the original actions did. Note: the two functions are different!
• MalformedOutputError:
  • Overall, the main reason of MalformedOutputError is there are some unexpected messages printed into the console, the messages may from the print function, or even from errors or warnings from pddl solver. Different details lead to different problems.
  • Based on the experiences from our classmates, we suggest:
    • Firstly, make sure that your codes can pass your local test before submission.
    • Double-check there is no other print function introduced except the only one from our template.
    • For crossing task, make sure you have modified the generatePlan() function correctly to add new actions. As this function is only called on AIVLE, so the error from this function doesn't appear in your local test.
    • For the WARNING: overflow on h^add! Costs clamped to 100000000:
      • The main reason is that the planner searched a very large space that make it overflow, especially on AIVLE, as our test cases are much more complex than your local test cases.
      • Please try to reduce your state space, PRECOND in action shames, and even goal state.
      • Especially, please don't add states about the agent blocked XXX cell while it's moving, we found these redundant states lead to this error easily.