praveen-palanisamy/macad-agents

Agents code for Multi-Agent Connected Autonomous Driving (MACAD) described in the paper presented in the Machine Learning for Autonomous Driving Workshop at NeurIPS 2019:

macad-agents's Issues

ModuleNotFoundError: No module named 'macad_gym.envs.intersection.stop_sign_urban_intersection_3c'

Hi, @praveen-palanisamy.
I want to use your code but an error has occurred in my situation. 😢
So I want your help.

I'm sorry for my poor English. 🙇

Reproduction

Run:

docker run --gpus all -it -e CARLA_SERVER=/software/CarlaUE4.sh -v ${HOME}/src/github.com/praveen-palanisamy/macad-agents:/macad-agents -w /macad-agents/src macad-agents:latest python -m macad_agents.rllib.impala_multiagent_shared_weights.py

Then, this error has occurred:

Traceback (most recent call last):
  File "/opt/conda/lib/python3.6/runpy.py", line 183, in _run_module_as_main
    mod_name, mod_spec, code = _get_module_details(mod_name, _Error)
  File "/opt/conda/lib/python3.6/runpy.py", line 109, in _get_module_details
    __import__(pkg_name)
  File "/macad-agents/src/macad_agents/rllib/impala_multiagent_shared_weights.py", line 15, in <module>
    from macad_gym.envs.intersection.stop_sign_urban_intersection_3c import \
ModuleNotFoundError: No module named 'macad_gym.envs.intersection.stop_sign_urban_intersection_3c'
Killing live carla processes set()

Environment

• OS: Ubuntu 18.04.3 LTS
• Docker: 19.03.5
• CPU: AMD Ryzen 5 3600X 6-Core Processor
• GPU: GeForce RTX 2070 (Driver Version: 430.50)

`conda install numpy` won't finish forever

Discussed in #17

^{Originally posted by rsuwa October 7, 2021}

macad-agents/Dockerfile

Line 103 in b2726b3

RUN conda install -y numpy

$ docker build --rm -f macad-agents/Dockerfile -t macad-agents:latest .
...

Step 8/15 : RUN conda install -y numpy
 ---> Running in 7123a03249d8
Solving environment: ...working...

No matter how many hours I wait, it stops as described above.

confusion about Baseline_agent

Hello,
I run:
python baseline_agent.py by loading the environment jason file in the intersection environments:
https://github.com/praveen-palanisamy/macad-gym/tree/master/src/macad_gym/envs/intersection

But only one vehicle can be controlled.
I also tried to run the baseline_agent with the default configuration. It is actually a single-vehicle drive straight forward.

Here are my two confusions about the baseline agent:

1. How to control multi-agent?
   As I saw in the source code:
   for actor_id in actor_configs.keys(): vehicle_dict[actor_id] = env._actors[actor_id]
   I think it should be able to control multi-agent.

2. How to set the duration of the simulation?

If I want to let the agent run in a longer period(i.e. more time steps per episode), how to change the environment parameters.

Thanks and look forward to your reply

How to run it without using docker?

How to run it without using docker?

ModuleNotFoundError: No module named 'env'

In this script:
https://github.com/praveen-palanisamy/macad-agents/blob/master/src/macad_agents/a3c/continuous_A3C_tune.py
when i run python -m macad_agents.a3c.continuous_A3C_tune i cannot find the env module.

Macad agent scripts running out of resouces

Hi @praveen-palanisamy @SHITIANYU-hue,

I am trying to run an impala agent independent script to train multi-agents. My environment is working fine but all the scripts apparently create multiple Carla instances and it results in freezing my laptop.
I took the screenshot to show the last logs on the terminal for reference.

During this screenshot, around 3 to 4 windows get opened and it stucks my laptop unless I reboot.

Any help in this would be appreciated.
Thanks

How to run baseline agent?

I try to run the baseline agent according to this line,

python -m macad_agents.rllib.baseline_agent

But I found this error:

agents.baseline_agent
free(): invalid pointer
Aborted (core dumped)

Inputs

hello
Is the input size [168,168,3] or [84,84,3] in https://github.com/praveen-palanisamy/macad-agents/blob/master/src/macad_agents/rllib/impala_multiagent_shared_weights.py？
In addition, conv_filter is [[32, [8], 4], [64, [4], 2], [64, [3, 3], 1)]?

Impala agent Code Issue

Hi @praveen-palanisamy ,

I am facing trouble while using Macad-agent impala examples. Even if I lower the resources the code runs but the cars don't move at all in default settings and without any training or crash Carla makes several windows of the environment.

Is there any way to verify the training of multi-agents through simulation? This way I can re-run the weights and see the results of multi-agent cars.

Thanks

Ray error when running continuous A3C

Hello,

when I run:

python -m macad_agents.a3c.continuous_A3C_tune.

I found that there will be the following error:
Traceback (most recent call last):
File "/home/tianyushi/miniconda3/envs/macad-gym/lib/python3.6/runpy.py", line 193, in _run_module_as_main
"main", mod_spec)
File "/home/tianyushi/miniconda3/envs/macad-gym/lib/python3.6/runpy.py", line 85, in _run_code
exec(code, run_globals)
File "/home/tianyushi/code/macad-agents/src/macad_agents/a3c/continuous_A3C_tune.py", line 74, in
"UPDATE_GLOBAL_ITER": 5,
File "/home/tianyushi/miniconda3/envs/macad-gym/lib/python3.6/site-packages/ray/tune/tune.py", line 185, in run_experiments
raise TuneError("Trials did not complete", errored_trials)
ray.tune.error.TuneError: ('Trials did not complete', [continuous_A3C_carla_env_0])
Killing live carla processes set()

and the log file is :
error_2020-08-27_16-06-25.txt

`conda install numpy` won't finish forever

Discussed in #17

^{Originally posted by rsuwa October 7, 2021}

macad-agents/Dockerfile

Line 103 in b2726b3

RUN conda install -y numpy

$ docker build --rm -f macad-agents/Dockerfile -t macad-agents:latest .
...

Step 8/15 : RUN conda install -y numpy
 ---> Running in 7123a03249d8
Solving environment: ...working...

No matter how many hours I wait, it stops as described above.

I have the same problem when building the Docker image locally. I tried twice, it always got stuck in here:

Step 8/15 : RUN conda install -y numpy
 ---> Running in 0ae7ce97b684
Solving environment: ...working...

is the continuous_a3c code valid?

I've tried the continuous_A3C.py, there exists some problems.

Problem

1. Incorrect dictionary update

macad-agents/src/macad_agents/a3c/continuous_A3C.py

Lines 30 to 33 in b2726b3

	env_config = DEFAULT_MULTIENV_CONFIG
	config_update = update_scenarios_parameter(
	json.load(open("macad_agents/a3c/env_config.json")))
	env_config.update(config_update)

Use dict.update to update a dictionary will override pre-exists keys in the dict, such as:

(d1,d2) = ({"Country":{}},{"Country":{}})
d1['Country']['area'] = 960
d2['Country']['population'] = 14

d1.update(d2) # The 'area' key will be replaced by 'population'

This will cause the "fixed_delta_seconds" key to be lost, as a result macad-gym can't initialize.

2. Serialization Problem

macad-agents/src/macad_agents/a3c/continuous_A3C.py

Lines 145 to 154 in b2726b3

	class Worker(mp.Process):
	def __init__(self, gnet, opt, global_ep, global_ep_r, res_queue, name):
	super(Worker, self).__init__()
	self.name = 'w%i' % name
	self.g_ep, self.g_ep_r, self.res_queue = (global_ep, global_ep_r,
	res_queue)
	self.gnet, self.opt = gnet, opt
	self.lnet = Net(N_S.spaces[vehicle_name],
	N_A.spaces[vehicle_name]) # local network
	self.env = MultiCarlaEnv(env_config)

Putting environment inside Net is not a good idea, when mp.Process serialize this object, it will try to serialize the environment as well, resulting in "can't pickle pygame.Font object" error.

Training

Even though I fixed those problems, it still doesn't seem to work. The "Mean Reward Curve" doesn't tend to go up (nor the distance curve to go down), and I haven't achieved one success episode yet (3M steps, maybe it's not enough to draw a conclusion?).

I know that PPO and IMPALA are the recommended algorithms, but since A3C is available in the repo, I want to know if it actually works.