{ "cells": [ { "cell_type": "markdown", "id": "2546904a", "metadata": {}, "source": [ "# Gymnasium integration\n", "[Gymnasium](https://gymnasium.farama.org/) is an API standard for reinforcement learning with a diverse collection of reference environments. Stormvogel supports some integration with gymnasium. In particular, you can construct explicit models from the gymnasium environmnents under Gymnasium's [ToyText](https://gymnasium.farama.org/environments/toy_text/) (except Blackjack)." ] }, { "cell_type": "markdown", "id": "80d60d9b", "metadata": {}, "source": [ "## FrozenLake\n", "Let us create one of these environments, called FrozenLake.\n", "Our agent wants to get to the present. Currently, it just chooses a random action." ] }, { "cell_type": "code", "execution_count": 1, "id": "9aa5d75e", "metadata": { "execution": { "iopub.execute_input": "2026-04-16T05:27:14.557703Z", "iopub.status.busy": "2026-04-16T05:27:14.557530Z", "iopub.status.idle": "2026-04-16T05:27:14.995426Z", "shell.execute_reply": "2026-04-16T05:27:14.994774Z" } }, "outputs": [ { "data": { "text/html": [ "" ], "text/plain": [ "" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "import gymnasium as gym\n", "from stormvogel.extensions.gym_grid import *\n", "from stormvogel import *\n", "\n", "env = gym.make(\n", " \"FrozenLake-v1\", render_mode=\"rgb_array\", is_slippery=False\n", ") # Set `is_slippery=True` for stochastic behavior\n", "filename = gymnasium_render_model_gif(env, filename=\"ice1\")\n", "extensions.embed_gif(filename)" ] }, { "cell_type": "markdown", "id": "8efd9528", "metadata": {}, "source": [ "We can convert it into an explicit MDP as follows. Each state has a label that relates to the coordinates of the tile." ] }, { "cell_type": "code", "execution_count": 2, "id": "9fc0361b", "metadata": { "execution": { "iopub.execute_input": "2026-04-16T05:27:14.997436Z", "iopub.status.busy": "2026-04-16T05:27:14.997166Z", "iopub.status.idle": "2026-04-16T05:27:15.190202Z", "shell.execute_reply": "2026-04-16T05:27:15.189639Z" } }, "outputs": [ { "data": { "text/html": [ "\n", "\n", "\n", " \n", " Network\n", " \n", " \n", " \n", " \n", " \n", "
\n", " \n", " \n", " \n", "\n" ], "text/plain": [ "" ] }, "metadata": {}, "output_type": "display_data" }, { "data": { "text/plain": [ "" ] }, "execution_count": 2, "metadata": {}, "output_type": "execute_result" } ], "source": [ "sv_model = gymnasium_grid_to_stormvogel(env, GRID_ACTION_LABEL_MAP)\n", "# GRID_ACTION_LABEL_MAP = {0: \"←\", 1: \"↓\", 2: \"→\", 3: \"↑\", 4: \"pickup\", 5: \"dropoff\"} (map between gymnasium action ids and labels)\n", "show(sv_model)" ] }, { "cell_type": "markdown", "id": "aad967ec", "metadata": {}, "source": [ "Now, let's do some model checking to calculate a strategy to solve the puzzle. We will tell the model checker to maximize the probability of getting to the target state (the present)." ] }, { "cell_type": "code", "execution_count": 3, "id": "23e6b42e", "metadata": { "execution": { "iopub.execute_input": "2026-04-16T05:27:15.197857Z", "iopub.status.busy": "2026-04-16T05:27:15.197505Z", "iopub.status.idle": "2026-04-16T05:27:15.235973Z", "shell.execute_reply": "2026-04-16T05:27:15.235133Z" } }, "outputs": [ { "data": { "text/html": [ "\n", "\n", "\n", " \n", " Network\n", " \n", " \n", " \n", " \n", " \n", "
\n", " \n", " \n", " \n", "\n" ], "text/plain": [ "" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "res = model_checking(sv_model, 'Pmax=? [F \"target\"]')\n", "vis2 = show(sv_model, result=res)" ] }, { "cell_type": "markdown", "id": "aaa8ba6e", "metadata": {}, "source": [ "Let's highlight the path to see what the scheduler is doing." ] }, { "cell_type": "code", "execution_count": 4, "id": "868b4caf", "metadata": { "execution": { "iopub.execute_input": "2026-04-16T05:27:15.243815Z", "iopub.status.busy": "2026-04-16T05:27:15.243597Z", "iopub.status.idle": "2026-04-16T05:27:30.304018Z", "shell.execute_reply": "2026-04-16T05:27:30.303379Z" } }, "outputs": [], "source": [ "path = simulate_path(sv_model, scheduler=res.scheduler, steps=20)\n", "vis2.highlight_path(path, color=\"orange\")" ] }, { "cell_type": "markdown", "id": "0ebf4628", "metadata": {}, "source": [ "Alternatively, we can show what our scheduler does in the frozen lake environment itself." ] }, { "cell_type": "code", "execution_count": 5, "id": "80c5cf3c", "metadata": { "execution": { "iopub.execute_input": "2026-04-16T05:27:30.306034Z", "iopub.status.busy": "2026-04-16T05:27:30.305850Z", "iopub.status.idle": "2026-04-16T05:27:30.351967Z", "shell.execute_reply": "2026-04-16T05:27:30.351342Z" } }, "outputs": [ { "data": { "text/html": [ "" ], "text/plain": [ "" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "from stormvogel.extensions.gym_grid import *\n", "\n", "gs = to_gymnasium_scheduler(sv_model, res.scheduler, GRID_ACTION_LABEL_MAP)\n", "filename = gymnasium_render_model_gif(env, gs, filename=\"ice2\")\n", "extensions.embed_gif(filename)" ] }, { "cell_type": "markdown", "id": "91f58e3b", "metadata": {}, "source": [ "We can also define a function to act as the scheduler on the model and convert it to a gymnasium scheduler. This one just keeps going in a loop..." ] }, { "cell_type": "code", "execution_count": 6, "id": "0710f5bf", "metadata": { "execution": { "iopub.execute_input": "2026-04-16T05:27:30.353846Z", "iopub.status.busy": "2026-04-16T05:27:30.353667Z", "iopub.status.idle": "2026-04-16T05:27:30.653129Z", "shell.execute_reply": "2026-04-16T05:27:30.652483Z" } }, "outputs": [ { "data": { "text/html": [ "" ], "text/plain": [ "" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "from stormvogel.model import Action\n", "\n", "\n", "def my_scheduler(s: stormvogel.model.State):\n", " # \"←\" \"↓\" \"→\" \"↑\"\n", " if s.is_initial():\n", " return Action(\"→\")\n", " env_id = s.get_valuation(stormvogel.model.Variable(\"env_id\"))\n", " x, y = to_coordinate(env_id, env)\n", " if x < 2 and y == 0:\n", " return Action(\"→\")\n", " elif x == 2 and y < 2:\n", " return Action(\"↓\")\n", " elif x > 0 and y == 2:\n", " return Action(\"←\")\n", " else:\n", " return Action(\"↑\")\n", "\n", "\n", "gs = to_gymnasium_scheduler(sv_model, my_scheduler, GRID_ACTION_LABEL_MAP)\n", "filename = gymnasium_render_model_gif(env, gs, filename=\"ice3\")\n", "extensions.embed_gif(filename)" ] }, { "cell_type": "markdown", "id": "cadeb305", "metadata": {}, "source": [ "## CliffWalking\n", "CliffWalking is a slightly more boring version of FrozenLake. You can apply the same principles that we just applied to FrozenLake." ] }, { "cell_type": "code", "execution_count": 7, "id": "bae11294", "metadata": { "execution": { "iopub.execute_input": "2026-04-16T05:27:30.655283Z", "iopub.status.busy": "2026-04-16T05:27:30.655087Z", "iopub.status.idle": "2026-04-16T05:27:30.661983Z", "shell.execute_reply": "2026-04-16T05:27:30.661360Z" } }, "outputs": [], "source": [ "import gymnasium as gym\n", "from stormvogel.extensions.gym_grid import *\n", "\n", "env = gym.make(\"CliffWalking-v1\", render_mode=\"rgb_array\")" ] }, { "cell_type": "code", "execution_count": 8, "id": "82a1a462", "metadata": { "execution": { "iopub.execute_input": "2026-04-16T05:27:30.663760Z", "iopub.status.busy": "2026-04-16T05:27:30.663576Z", "iopub.status.idle": "2026-04-16T05:27:30.703044Z", "shell.execute_reply": "2026-04-16T05:27:30.702503Z" } }, "outputs": [ { "data": { "text/html": [ "\n", "\n", "\n", " \n", " Network\n", " \n", " \n", " \n", " \n", " \n", "
\n", " \n", " \n", " \n", "\n" ], "text/plain": [ "" ] }, "metadata": {}, "output_type": "display_data" }, { "data": { "text/plain": [ "" ] }, "execution_count": 8, "metadata": {}, "output_type": "execute_result" } ], "source": [ "sv_model = gymnasium_grid_to_stormvogel(env, GRID_ACTION_LABEL_MAP)\n", "show(sv_model)" ] }, { "cell_type": "code", "execution_count": 9, "id": "6b3eab59", "metadata": { "execution": { "iopub.execute_input": "2026-04-16T05:27:30.710412Z", "iopub.status.busy": "2026-04-16T05:27:30.710204Z", "iopub.status.idle": "2026-04-16T05:27:30.903065Z", "shell.execute_reply": "2026-04-16T05:27:30.902408Z" } }, "outputs": [ { "data": { "text/html": [ "" ], "text/plain": [ "" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "from stormvogel.stormpy_utils.model_checking import model_checking\n", "\n", "res = model_checking(sv_model, 'Pmax=? [F \"target\"]')\n", "gs = to_gymnasium_scheduler(sv_model, res.scheduler, GRID_ACTION_LABEL_MAP)\n", "filename = gymnasium_render_model_gif(env, gs, filename=\"cliff\")\n", "extensions.embed_gif(filename)" ] }, { "cell_type": "markdown", "id": "d33d7ea5", "metadata": {}, "source": [ "## Taxi\n", "In the Taxi scenario, a taxi has to pick up passengers and transport them to the hotel. The position of the target, passenger and taxi are chosen at random." ] }, { "cell_type": "code", "execution_count": 10, "id": "048b65ba", "metadata": { "execution": { "iopub.execute_input": "2026-04-16T05:27:30.904966Z", "iopub.status.busy": "2026-04-16T05:27:30.904767Z", "iopub.status.idle": "2026-04-16T05:27:31.115190Z", "shell.execute_reply": "2026-04-16T05:27:31.114542Z" } }, "outputs": [ { "data": { "text/plain": [ "'ModelType.MDP model with 3529 states, 6049 choices, and 1005 distinct labels.'" ] }, "execution_count": 10, "metadata": {}, "output_type": "execute_result" } ], "source": [ "import gymnasium as gym\n", "from stormvogel.extensions.gym_grid import *\n", "\n", "env = gym.make(\n", " \"Taxi-v3\", render_mode=\"rgb_array\"\n", ") # Set `is_slippery=True` for stochastic behavior\n", "sv_model = gymnasium_grid_to_stormvogel(env)\n", "# This model is so big that it is better not to display it.\n", "sv_model.summary()" ] }, { "cell_type": "code", "execution_count": 11, "id": "5de9fb9d", "metadata": { "execution": { "iopub.execute_input": "2026-04-16T05:27:31.117021Z", "iopub.status.busy": "2026-04-16T05:27:31.116830Z", "iopub.status.idle": "2026-04-16T05:27:32.001873Z", "shell.execute_reply": "2026-04-16T05:27:32.001285Z" } }, "outputs": [ { "data": { "text/html": [ "" ], "text/plain": [ "" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "target = get_target_state(env)\n", "res = model_checking(sv_model, \"Rmax=? [S]\")\n", "gs = to_gymnasium_scheduler(sv_model, res.scheduler, GRID_ACTION_LABEL_MAP)\n", "filename = gymnasium_render_model_gif(env, gs, filename=\"taxi\")\n", "extensions.embed_gif(filename)" ] } ], "metadata": { "kernelspec": { "display_name": "Python 3 (ipykernel)", "language": "python", "name": "python3" }, "language_info": { "codemirror_mode": { "name": "ipython", "version": 3 }, "file_extension": ".py", "mimetype": "text/x-python", "name": "python", "nbconvert_exporter": "python", "pygments_lexer": "ipython3", "version": "3.12.13" }, "widgets": { "application/vnd.jupyter.widget-state+json": { "state": { "0e06503d6c984576a56c485712754623": { "model_module": "@jupyter-widgets/base", "model_module_version": "2.0.0", "model_name": "LayoutModel", "state": { "_model_module": "@jupyter-widgets/base", "_model_module_version": "2.0.0", "_model_name": "LayoutModel", "_view_count": null, "_view_module": "@jupyter-widgets/base", "_view_module_version": "2.0.0", "_view_name": "LayoutView", "align_content": null, "align_items": null, "align_self": null, "border_bottom": null, "border_left": null, "border_right": null, "border_top": null, "bottom": null, "display": null, "flex": null, "flex_flow": null, "grid_area": null, "grid_auto_columns": null, "grid_auto_flow": null, "grid_auto_rows": null, "grid_column": null, "grid_gap": null, "grid_row": null, "grid_template_areas": null, "grid_template_columns": null, "grid_template_rows": null, "height": null, "justify_content": null, "justify_items": null, "left": null, "margin": null, "max_height": null, "max_width": null, "min_height": null, "min_width": null, "object_fit": null, "object_position": null, "order": null, "overflow": null, "padding": null, "right": null, "top": null, "visibility": null, "width": null } }, "16f449f010e841909cff71e0d12b72a8": { "model_module": "@jupyter-widgets/output", "model_module_version": "1.0.0", "model_name": "OutputModel", "state": { "_dom_classes": [], "_model_module": "@jupyter-widgets/output", "_model_module_version": "1.0.0", "_model_name": "OutputModel", "_view_count": null, "_view_module": "@jupyter-widgets/output", "_view_module_version": "1.0.0", "_view_name": "OutputView", "layout": "IPY_MODEL_0e06503d6c984576a56c485712754623", "msg_id": "", "outputs": [], "tabbable": null, "tooltip": null } }, "3849cdf8e8844e0490c02839594a5a14": { "model_module": "@jupyter-widgets/base", "model_module_version": "2.0.0", "model_name": "LayoutModel", "state": { "_model_module": "@jupyter-widgets/base", "_model_module_version": "2.0.0", "_model_name": "LayoutModel", "_view_count": null, "_view_module": "@jupyter-widgets/base", "_view_module_version": "2.0.0", "_view_name": "LayoutView", "align_content": null, "align_items": null, "align_self": null, "border_bottom": null, "border_left": null, "border_right": null, "border_top": null, "bottom": null, "display": null, "flex": null, "flex_flow": null, "grid_area": null, "grid_auto_columns": null, "grid_auto_flow": null, "grid_auto_rows": null, "grid_column": null, "grid_gap": null, "grid_row": null, "grid_template_areas": null, "grid_template_columns": null, "grid_template_rows": null, "height": null, "justify_content": null, "justify_items": null, "left": null, "margin": null, "max_height": null, "max_width": null, "min_height": null, "min_width": null, "object_fit": null, "object_position": null, "order": null, "overflow": null, "padding": null, "right": null, "top": null, "visibility": null, "width": null } }, "45cf06e8bc0644e7ba3131a5d1374ced": { "model_module": "@jupyter-widgets/output", "model_module_version": "1.0.0", "model_name": "OutputModel", "state": { "_dom_classes": [], "_model_module": "@jupyter-widgets/output", "_model_module_version": "1.0.0", "_model_name": "OutputModel", "_view_count": null, "_view_module": "@jupyter-widgets/output", "_view_module_version": "1.0.0", "_view_name": "OutputView", "layout": "IPY_MODEL_3849cdf8e8844e0490c02839594a5a14", "msg_id": "", "outputs": [], "tabbable": null, "tooltip": null } }, "49f6d2c382eb4543ae596e8668383329": { "model_module": "@jupyter-widgets/base", "model_module_version": "2.0.0", "model_name": "LayoutModel", "state": { "_model_module": "@jupyter-widgets/base", "_model_module_version": "2.0.0", "_model_name": "LayoutModel", "_view_count": null, "_view_module": "@jupyter-widgets/base", "_view_module_version": "2.0.0", "_view_name": "LayoutView", "align_content": null, "align_items": null, "align_self": null, "border_bottom": null, "border_left": null, "border_right": null, "border_top": null, "bottom": null, "display": null, "flex": null, "flex_flow": null, "grid_area": null, "grid_auto_columns": null, "grid_auto_flow": null, "grid_auto_rows": null, "grid_column": null, "grid_gap": null, "grid_row": null, "grid_template_areas": null, "grid_template_columns": null, "grid_template_rows": null, "height": null, "justify_content": null, "justify_items": null, "left": null, "margin": null, "max_height": null, "max_width": null, "min_height": null, "min_width": null, "object_fit": null, "object_position": null, "order": null, "overflow": null, "padding": null, "right": null, "top": null, "visibility": null, "width": null } }, "4ea8c717d4354769858aee43fcaa8825": { "model_module": "@jupyter-widgets/base", "model_module_version": "2.0.0", "model_name": "LayoutModel", "state": { "_model_module": "@jupyter-widgets/base", "_model_module_version": "2.0.0", "_model_name": "LayoutModel", "_view_count": null, "_view_module": "@jupyter-widgets/base", "_view_module_version": "2.0.0", "_view_name": "LayoutView", "align_content": null, "align_items": null, "align_self": null, "border_bottom": null, "border_left": null, "border_right": null, "border_top": null, "bottom": null, "display": null, "flex": null, "flex_flow": null, "grid_area": null, "grid_auto_columns": null, "grid_auto_flow": null, "grid_auto_rows": null, "grid_column": null, "grid_gap": null, "grid_row": null, "grid_template_areas": null, "grid_template_columns": null, "grid_template_rows": null, "height": null, "justify_content": null, "justify_items": null, "left": null, "margin": null, "max_height": null, "max_width": null, "min_height": null, "min_width": null, "object_fit": null, "object_position": null, "order": null, "overflow": null, "padding": null, "right": null, "top": null, "visibility": null, "width": null } }, "6a52dd96dbf84448975fd8d747cd92f2": { "model_module": "@jupyter-widgets/output", "model_module_version": "1.0.0", "model_name": "OutputModel", "state": { "_dom_classes": [], "_model_module": "@jupyter-widgets/output", "_model_module_version": "1.0.0", "_model_name": "OutputModel", "_view_count": null, "_view_module": "@jupyter-widgets/output", "_view_module_version": "1.0.0", "_view_name": "OutputView", "layout": "IPY_MODEL_cf9e723614474106a94365f9abb0bc67", "msg_id": "", "outputs": [], "tabbable": null, "tooltip": null } }, "8bb62b8c35a748428e18d1cafb88963f": { "model_module": "@jupyter-widgets/base", "model_module_version": "2.0.0", "model_name": "LayoutModel", "state": { "_model_module": "@jupyter-widgets/base", "_model_module_version": "2.0.0", "_model_name": "LayoutModel", "_view_count": null, "_view_module": "@jupyter-widgets/base", "_view_module_version": "2.0.0", "_view_name": "LayoutView", "align_content": null, "align_items": null, "align_self": null, "border_bottom": null, "border_left": null, "border_right": null, "border_top": null, "bottom": null, "display": null, "flex": null, "flex_flow": null, "grid_area": null, "grid_auto_columns": null, "grid_auto_flow": null, "grid_auto_rows": null, "grid_column": null, "grid_gap": null, "grid_row": null, "grid_template_areas": null, "grid_template_columns": null, "grid_template_rows": null, "height": null, "justify_content": null, "justify_items": null, "left": null, "margin": null, "max_height": null, "max_width": null, "min_height": null, "min_width": null, "object_fit": null, "object_position": null, "order": null, "overflow": null, "padding": null, "right": null, "top": null, "visibility": null, "width": null } }, "b8d9012a1e3349c0a66b403e060115b6": { "model_module": "@jupyter-widgets/output", "model_module_version": "1.0.0", "model_name": "OutputModel", "state": { "_dom_classes": [], "_model_module": "@jupyter-widgets/output", "_model_module_version": "1.0.0", "_model_name": "OutputModel", "_view_count": null, "_view_module": "@jupyter-widgets/output", "_view_module_version": "1.0.0", "_view_name": "OutputView", "layout": "IPY_MODEL_4ea8c717d4354769858aee43fcaa8825", "msg_id": "", "outputs": [], "tabbable": null, "tooltip": null } }, "cc723d956dd24d7aa50d9161d618f7d0": { "model_module": "@jupyter-widgets/output", "model_module_version": "1.0.0", "model_name": "OutputModel", "state": { "_dom_classes": [], "_model_module": "@jupyter-widgets/output", "_model_module_version": "1.0.0", "_model_name": "OutputModel", "_view_count": null, "_view_module": "@jupyter-widgets/output", "_view_module_version": "1.0.0", "_view_name": "OutputView", "layout": "IPY_MODEL_8bb62b8c35a748428e18d1cafb88963f", "msg_id": "", "outputs": [], "tabbable": null, "tooltip": null } }, "cf9e723614474106a94365f9abb0bc67": { "model_module": "@jupyter-widgets/base", "model_module_version": "2.0.0", "model_name": "LayoutModel", "state": { "_model_module": "@jupyter-widgets/base", "_model_module_version": "2.0.0", "_model_name": "LayoutModel", "_view_count": null, "_view_module": "@jupyter-widgets/base", "_view_module_version": "2.0.0", "_view_name": "LayoutView", "align_content": null, "align_items": null, "align_self": null, "border_bottom": null, "border_left": null, "border_right": null, "border_top": null, "bottom": null, "display": null, "flex": null, "flex_flow": null, "grid_area": null, "grid_auto_columns": null, "grid_auto_flow": null, "grid_auto_rows": null, "grid_column": null, "grid_gap": null, "grid_row": null, "grid_template_areas": null, "grid_template_columns": null, "grid_template_rows": null, "height": null, "justify_content": null, "justify_items": null, "left": null, "margin": null, "max_height": null, "max_width": null, "min_height": null, "min_width": null, "object_fit": null, "object_position": null, "order": null, "overflow": null, "padding": null, "right": null, "top": null, "visibility": null, "width": null } }, "d13e309321af47f283249be615571c49": { "model_module": "@jupyter-widgets/output", "model_module_version": "1.0.0", "model_name": "OutputModel", "state": { "_dom_classes": [], "_model_module": "@jupyter-widgets/output", "_model_module_version": "1.0.0", "_model_name": "OutputModel", "_view_count": null, "_view_module": "@jupyter-widgets/output", "_view_module_version": "1.0.0", "_view_name": "OutputView", "layout": "IPY_MODEL_49f6d2c382eb4543ae596e8668383329", "msg_id": "", "outputs": [], "tabbable": null, "tooltip": null } } }, "version_major": 2, "version_minor": 0 } } }, "nbformat": 4, "nbformat_minor": 5 }