First tests done

README.md

@@ -45,13 +45,13 @@ This repository contains the implementation of our paper [SocNavGym: A Reinforce
 ## Usage
 ```python
 import socnavgym
-import gym
+import gymnasium as gym
 env = gym.make('SocNavGym-v1', config="<PATH_TO_CONFIG>")  
 ```
 ## Sample Code
 ```python
 import socnavgym
-import gym
+import gymnasium as gym
 env = gym.make("SocNavGym-v1", config="./environment_configs/exp1_no_sngnn.yaml") 
 obs, _ = env.reset()
 
@@ -64,7 +64,7 @@ for i in range(1000):
 ```
 
 ## About the environment
-```SocNavGym-v1``` is a highly customisable environment and the parameters of the environment can be controlled using the config files. There are a few config files in [environment_configs/](https://github.com/gnns4hri/SocNavGym/tree/main/environment_configs). For a better understanding of each parameter refer to [this](https://github.com/gnns4hri/SocNavGym#config-file) section. Other than the robot, the environment supports entities like plants, tables, laptops. The environment also models interactions between humans, and human-laptop. It can also contain moving crowds, and static crowds, and the ability to form new crowds and interactions, as well as disperse existing crowds and interactions. The environment follows the OpenAI Gym format implementing the `step`, `render` and `reset` functions. The environment uses the latest Gym API (gym 0.26.2).
+```SocNavGym-v1``` is a highly customisable environment and the parameters of the environment can be controlled using the config files. There are a few config files in [environment_configs/](https://github.com/gnns4hri/SocNavGym/tree/main/environment_configs). For a better understanding of each parameter refer to [this](https://github.com/gnns4hri/SocNavGym#config-file) section. Other than the robot, the environment supports entities like plants, tables, laptops. The environment also models interactions between humans, and human-laptop. It can also contain moving crowds, and static crowds, and the ability to form new crowds and interactions, as well as disperse existing crowds and interactions. The environment follows the Farama foundation format implementing the `step`, `render` and `reset` functions. The environment uses the latest Gymnasium API (gymnasium 0.29.1).
 
 ## Conventions
 * X-axis points in the direction of zero-angle.
@@ -74,7 +74,7 @@ for i in range(1000):
 ## Observation Space
 The observation returned when ```env.step(action)``` is called, consists of the following (all in the<b> robot frame</b> unless you're using the [`WorldFrameObservations`](https://github.com/gnns4hri/SocNavGym#wrappers) wrapper):
 
-The observation is of the type `gym.Spaces.Dict`. The dictionary has the following keys:
+The observation is of the type `gymnasium.Spaces.Dict`. The dictionary has the following keys:
 1. ```"robot"``` : This is a vector of shape (9,) of which the first six values represent the one-hot encoding of the robot, i.e ```[1, 0, 0, 0, 0, 0]```. The next two values represent the goal's x and y coordinates in the robot frame and the last value is the robot's radius.
 
 2. The other keys present in the observation are ```"humans"```, ```"plants"```, ```"laptops"```, ```"tables"``` and ```"walls"```. Every entity (human, plant, laptop, table, or wall) would have an observation vector given by the structure below:
@@ -628,7 +628,7 @@ The behaviour of the environment is controlled using the config file. The config
 </table>
 
 ## Wrappers
-Gym wrappers are convenient to have changes in the observation-space / action-space. SocNavGym implements 4 wrappers. 
+Gymnasium wrappers are convenient to have changes in the observation-space / action-space. SocNavGym implements 4 wrappers. 
 
 The following are the wrappers implemented by SocNavGym:
 1. `DiscreteActions` : To change the environment from a continuous action space environment to a discrete action space environment. The action space consists of 7 discrete actions. They are :

manual_control_js.py

@@ -1,6 +1,6 @@
 import torch
 import time
-import gym
+import gymnasium as gym
 import numpy as np
 import socnavgym
 import os

sb3_eval.py

@@ -1,4 +1,4 @@
-import gym
+import gymnasium as gym
 import socnavgym
 import torch
 from socnavgym.wrappers import DiscreteActions

setup.py

@@ -16,7 +16,7 @@
     LONG_DESCRIPTION = f.read()
 
 REQUIRED = [
-    "gym >= 0.26.2",
+    "gymnasium >= 0.29.1",
     "opencv-python",
     "numpy",
     "matplotlib",

socnavgym/__init__.py

@@ -1,4 +1,4 @@
-from gym.envs.registration import register
+from gymnasium.envs.registration import register
 
 register(
     id='SocNavGym-v1',

socnavgym/envs/socnavenv_v1.py

@@ -7,13 +7,13 @@
 import copy
 from importlib.machinery import SourceFileLoader
 import cv2
-import gym
+import gymnasium as gym
 import matplotlib.pyplot as plt
 import numpy as np
 import rvo2
 import torch
 import yaml
-from gym import spaces
+from gymnasium import spaces
 from shapely.geometry import Point, Polygon
 from collections import namedtuple
 import math
@@ -146,14 +146,19 @@ class SocNavEnv_v1(gym.Env):
     # human-laptop interaction params
     HUMAN_LAPTOP_DISTANCE = None
 
-    def __init__(self, config:str=None) -> None:
+    def __init__(self, config:str=None, render_mode:str=None) -> None:
         """
         Args : 
             config: Path to the environment config file
         """
         super().__init__()
 
-        assert(config is not None), "Argument config_path is None. Please call gym.make(\"SocNavGym-v1\", config_path=path_to_config)"
+        assert(config is not None), "Argument config is None. Please call gym.make(\"SocNavGym-v1\", config=path_to_config)"
+
+
+        self.render_mode_ = render_mode
+        if render_mode is None:
+            self.render_mode_ = "human"
 
         self.window_initialised = False
         self.has_configured = False
@@ -3605,12 +3610,12 @@ def render_without_showing(self, mode="human", draw_human_gaze=False, draw_human
 
         return self.world_image
 
-    def render(self, mode="human", draw_human_gaze=False, draw_human_goal=True):
+    def render(self, draw_human_gaze=False, draw_human_goal=True):
         if not self.window_initialised:
             cv2.namedWindow("world", cv2.WINDOW_NORMAL) 
             cv2.resizeWindow("world", int(self.RESOLUTION_VIEW), int(self.RESOLUTION_VIEW))
             self.window_initialised = True
-        self.world_image = self.render_without_showing(mode, draw_human_gaze, draw_human_goal)
+        self.world_image = self.render_without_showing(self.render_mode_, draw_human_gaze, draw_human_goal)
         cv2.imshow("world", self.world_image)
         k = cv2.waitKey(self.MILLISECONDS)
         if k%255 == 27:

socnavgym/wrappers/discrete_actions.py

@@ -1,5 +1,5 @@
-import gym
-from gym.spaces import Discrete
+import gymnasium as gym
+from gymnasium.spaces import Discrete
 from socnavgym.envs.socnavenv_v1 import SocNavEnv_v1
 import numpy as np
 

socnavgym/wrappers/noisy_observations.py

@@ -1,5 +1,5 @@
-import gym
-from gym import spaces
+import gymnasium as gym
+from gymnasium import spaces
 from socnavgym.envs.socnavenv_v1 import SocNavEnv_v1
 from socnavgym.envs.utils.wall import Wall
 import numpy as np

socnavgym/wrappers/partial_observations.py

@@ -1,5 +1,5 @@
-import gym
-from gym import spaces
+import gymnasium as gym
+from gymnasium import spaces
 from socnavgym.envs.socnavenv_v1 import SocNavEnv_v1
 from socnavgym.envs.utils.wall import Wall
 from socnavgym.envs.utils.utils import w2px, w2py

socnavgym/wrappers/world_frame_observations.py

@@ -1,5 +1,5 @@
-import gym
-from gym import spaces
+import gymnasium as gym
+from gymnasium import spaces
 from socnavgym.envs.socnavenv_v1 import SocNavEnv_v1
 from socnavgym.envs.utils.wall import Wall
 import numpy as np

stable_dqn.py

@@ -1,4 +1,4 @@
-import gym
+import gymnasium as gym
 import socnavgym
 import torch
 import torch.nn as nn

tests/test_configs.py

@@ -2,10 +2,10 @@
 import sys
 import os
 sys.path.insert(1, os.path.dirname(os.path.abspath(__file__)) + "/..")
-import gym
+import gymnasium as gym
 import socnavgym
 from socnavgym.wrappers import WorldFrameObservations, PartialObservations, NoisyObservations, DiscreteActions
-from gym.utils.env_checker import check_env
+from gymnasium.utils.env_checker import check_env
 import numpy as np
 from glob import glob
 

tests/test_env.py

@@ -2,10 +2,10 @@
 import sys
 import os
 sys.path.insert(1, os.path.dirname(os.path.abspath(__file__)) + "/..")
-import gym
+import gymnasium as gym
 import socnavgym
 from socnavgym.wrappers import WorldFrameObservations, PartialObservations, NoisyObservations, DiscreteActions
-from gym.utils.env_checker import check_env
+from gymnasium.utils.env_checker import check_env
 import numpy as np
 
 

tests/test_observations.py

@@ -2,7 +2,7 @@
 import sys
 import os
 sys.path.insert(1, os.path.dirname(os.path.abspath(__file__)) + "/..")
-import gym
+import gymnasium as gym
 import socnavgym
 import yaml
 import numpy as np

tests/test_shape.py

@@ -2,7 +2,7 @@
 import sys
 import os
 sys.path.insert(1, os.path.dirname(os.path.abspath(__file__)) + "/..")
-import gym
+import gymnasium as gym
 import socnavgym
 import yaml