OfflineSimulation.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "171f3af0",
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "import tensorflow as tf\n",
    "from tensorflow import keras\n",
    "from matplotlib import pyplot as plt\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "from scipy.linalg import expm\n",
    "from copy import deepcopy\n",
    "import xgboost as xgb\n",
    "import time\n",
    "from sklearn.metrics import classification_report,average_precision_score, precision_recall_curve, roc_curve, roc_auc_score, f1_score, confusion_matrix, auc, precision_score, recall_score\n",
    "import seaborn as sns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "54e22a55",
   "metadata": {},
   "outputs": [],
   "source": [
    "# model = xgb.XGBClassifier(objective = 'binary:logistic', \n",
    "#                         seed=42, \n",
    "#                         n_estimators=230,\n",
    "#                         eval_metric='aucpr',\n",
    "#                         n_jobs = -1)\n",
    "\n",
    "# model.load_model(\"xgboost_sim_full.json\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "37bf334d",
   "metadata": {},
   "outputs": [],
   "source": [
    "# results = model.evals_result()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "a3234453",
   "metadata": {},
   "outputs": [],
   "source": [
    "df = pd.read_csv('./robotdatacollection3_part2/ep1.csv')\n",
    "# df = pd.read_csv('./robot_data/ep4_new.csv')\n",
    "\n",
    "# model = keras.models.load_model('./sim_final_full')\n",
    "model = keras.models.load_model('./Thesis/thesis2_classifier')\n",
    "\n",
    "# model = keras.models.load_model('./robot_1_full')\n",
    "robot_flag=True\n",
    "\n",
    "threshold = 0.5"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "f9e7680d",
   "metadata": {},
   "outputs": [],
   "source": [
    "ERROR_TOP = 0.0008\n",
    "PEG_RADIUS = 0.0021\n",
    "HOLE_RADIUS = 0.0024\n",
    "\n",
    "MEMORY_LEN = 100\n",
    "FEATURE_SIZE = 5\n",
    "cnt = 0\n",
    "robot = True"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "41848970",
   "metadata": {},
   "outputs": [],
   "source": [
    "memory = [[0] * FEATURE_SIZE for _ in range(MEMORY_LEN)]\n",
    "pred_num = 5\n",
    "pred_memory = [0 for _ in range(pred_num)]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "a7a732eb",
   "metadata": {},
   "outputs": [],
   "source": [
    "feature_names = ['Fx', 'Fy', 'Fz', 'Mx', 'My']\n",
    "# feature_names = ['fz']\n",
    "# \n",
    "# feature_names = ['fx', 'fy', 'fz', 'mx', 'my']\n",
    "\n",
    "X_df = df[feature_names]\n",
    "y = df.Case.to_numpy()\n",
    "\n",
    "# moving average\n",
    "# roll = 30\n",
    "# X_df = X_df.rolling(roll).mean()\n",
    "# X_df = X_df.loc[roll-1:]\n",
    "# y = y[roll-1:]\n",
    "\n",
    "\n",
    "X = X_df.to_numpy()\n",
    "\n",
    "window_len = 100#300\n",
    "\n",
    "if window_len:\n",
    "    n_features = len(feature_names)  # 3\n",
    "\n",
    "    row = X.shape[0] + 1 - window_len\n",
    "    col_len = n_features * window_len\n",
    "    new_x = np.zeros((row, col_len))\n",
    "    new_y = np.zeros((row, 1))\n",
    "\n",
    "    for i in range(len(new_x)):\n",
    "        new_x[i] = X[i:i + window_len].reshape(1, col_len)[0][::-1]\n",
    "        new_y[i] = y[i + (window_len - 1)]\n",
    "\n",
    "    y = deepcopy(new_y)\n",
    "    X = deepcopy(new_x)\n",
    "\n",
    "total = len(y)\n",
    "pos = int(y.sum())\n",
    "neg = total - pos\n",
    "weight = neg/pos # 13.5"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "07a8b879",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "170/170 [==============================] - 0s 1ms/step\n"
     ]
    }
   ],
   "source": [
    "y_test_pred = model.predict(X)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "822c8015",
   "metadata": {},
   "outputs": [],
   "source": [
    "def fancy_confusion(cf_matrix, title):\n",
    "    group_counts = ['{0:0.0f}'.format(value) for value in\n",
    "                    cf_matrix.flatten()]\n",
    "    group_percentages = ['{0:.2%}'.format(value) for value in\n",
    "                         cf_matrix.flatten()/np.sum(cf_matrix)]\n",
    "\n",
    "    labels = [f\"{v1}\\n{v2}\" for v1, v2 in\n",
    "              zip(group_counts,group_percentages)]\n",
    "    labels = np.asarray(labels).reshape(2,2)\n",
    "\n",
    "    plt.figure()\n",
    "    plt.title(title)\n",
    "    sns.heatmap(cf_matrix, annot=labels,cmap='YlOrRd',fmt=\"\")\n",
    "    plt.xlabel('Predicted Class')\n",
    "    plt.ylabel('Actual Class')\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "81be217b",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Perfect confusion matrix\n",
      "[[5310    0]\n",
      " [   0  125]]\n"
     ]
    }
   ],
   "source": [
    "print(\"Perfect confusion matrix\")\n",
    "print(confusion_matrix(y, y > 0.5))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "edd05e05",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 640x480 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fancy_confusion(confusion_matrix(y, y > 0.5),\"Perfect confusion matrix\" )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "id": "bec60386",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[5302    8]\n",
      " [ 125    0]]\n"
     ]
    }
   ],
   "source": [
    "print(confusion_matrix(y, y_test_pred > 0.5))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "id": "dc93d041",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 640x480 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fancy_confusion(confusion_matrix(y, y_test_pred > 0.5),\"Test confusion matrix\" )"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a81cbe56",
   "metadata": {},
   "source": [
    "### Stream sensor measuerements and let it predict - sensor measurements are already after contact!\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "id": "46fd627c",
   "metadata": {},
   "outputs": [],
   "source": [
    "def append_vector(array, vector):\n",
    "    # Discard the oldest value\n",
    "    array.pop()\n",
    "    # Add the new value\n",
    "    array.insert(0, vector)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "4ad7a69e",
   "metadata": {},
   "outputs": [],
   "source": [
    " def circle_check(peg_x,peg_y,robot):\n",
    "        # CHANGE FOR SIM AND ROBOT\n",
    "        if robot:\n",
    "            hole_x = -0.1379\n",
    "            hole_y = -0.5095\n",
    "        else:\n",
    "            hole_x = 0\n",
    "            hole_y = 0\n",
    "\n",
    "        print(np.sqrt((peg_x - hole_x) ** 2 + (peg_y - hole_y) ** 2))\n",
    "        \n",
    "        if np.sqrt((peg_x - hole_x) ** 2 + (peg_y - hole_y) ** 2) < ERROR_TOP:\n",
    "            return True\n",
    "        else:\n",
    "            return False\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "78659b8c",
   "metadata": {},
   "outputs": [],
   "source": [
    "df.columns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "e1d29f1b",
   "metadata": {},
   "outputs": [],
   "source": [
    "x = df.x.values\n",
    "y = df.y.values\n",
    "z = df.z.values\n",
    "# fx = df.fx.values\n",
    "# fy = df.fy.values\n",
    "# fz = df.fz.values\n",
    "# mx = df.mx.values\n",
    "# my = df.my.values\n",
    "# mz = df.mz.values\n",
    "# case = df.case.values\n",
    "\n",
    "fx = df.Fx.values\n",
    "fy = df.Fy.values\n",
    "fz = df.Fz.values\n",
    "mx = df.Mx.values\n",
    "my = df.My.values\n",
    "mz = df.Mz.values\n",
    "case = df.Case.values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "dae43a81",
   "metadata": {},
   "outputs": [],
   "source": [
    "start_idx = df.index[df['Case'].diff() == 1].tolist()\n",
    "print(start_idx)\n",
    "print(df.loc[start_idx[0]].t)\n",
    "print(df.loc[start_idx[1]].t)\n",
    "# print(df.loc[start_idx[2]].t)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "3035d13c",
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# DNN\n",
    "vnt=0\n",
    "switch = False\n",
    "cnt_list = []\n",
    "while cnt < len(df):\n",
    "    # fx,fy,fz,mx,my\n",
    "    \n",
    "    features = np.array([fx[cnt],fy[cnt],fz[cnt],mx[cnt],my[cnt]])\n",
    "    rev_features = features[::-1]  # reverse to match inputs for training NN\n",
    "    features_list = rev_features.tolist()\n",
    "    \n",
    "    append_vector(memory, features_list)\n",
    "    # we need to fill out the buffer else it's filled with zeros\n",
    "    if cnt >= MEMORY_LEN:\n",
    "        X_vec = np.array(memory).reshape(1, MEMORY_LEN*FEATURE_SIZE)\n",
    "#         print(X_vec.shape)\n",
    "        t_init=time.time()\n",
    "        y_predict = model(X_vec)\n",
    "        print('PRED_TIME',time.time()-t_init)\n",
    "        temp_overlap = (y_predict >= threshold)\n",
    "        \n",
    "        print('prediction', y_predict)\n",
    "        inst_overlap = temp_overlap.numpy()[0][0]  # boolean true/false conversion tensor->bool\n",
    "        print('Model detected overlap', inst_overlap)\n",
    "        print('Geometric overlap', circle_check(df.x.values[cnt],df.y.values[cnt],robot=robot_flag))\n",
    "        print('Time', df.t.values[cnt])\n",
    "        print(cnt)\n",
    "        print()\n",
    "        append_vector(pred_memory, inst_overlap)\n",
    "        if all(pred_memory):\n",
    "            print('SWITCH!!!!')\n",
    "            switch = True\n",
    "            cnt_list.append(cnt)\n",
    "#             cnt = len(df)-1\n",
    "    cnt += 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "74ea8e51",
   "metadata": {},
   "outputs": [],
   "source": [
    "one_occurace = np.where(df.Case == 1)[0]\n",
    "print(one_occurace)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "189290c6",
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "cnt_list"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "83262827",
   "metadata": {},
   "outputs": [],
   "source": [
    "def plot_with_labels(df, name, show_label):\n",
    "    plt.figure(figsize=(10,4))\n",
    "    plt.title(name)\n",
    "    plt.plot(df.t, df[name])\n",
    "    if show_label:\n",
    "        for i in cnt_list:\n",
    "            plt.scatter(df.t.loc[i], df[name].loc[i],marker='*', color='red', zorder=12, alpha=0.2)\n",
    "    \n",
    "    plt.grid()\n",
    "    plt.show()\n",
    "\n",
    "def plot_vs_time(df, name):\n",
    "    plt.figure(figsize=(10,4))\n",
    "    plt.title(name)\n",
    "    plt.plot(df.t, df[name])\n",
    "    plt.grid()\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "de79b526",
   "metadata": {},
   "outputs": [],
   "source": [
    "for f in feature_names:\n",
    "    plot_with_labels(df=df,name=f,show_label=True)\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "44ef7433",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "28697259",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "0965caab",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "6925718c",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "f1cee554",
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# # XGB\n",
    "\n",
    "# switch = False\n",
    "# while cnt < len(df):\n",
    "#     # fx,fy,fz,mx,my\n",
    "    \n",
    "#     features = np.array([fx[cnt],fy[cnt],fz[cnt],mx[cnt],my[cnt]])\n",
    "#     rev_features = features[::-1]  # reverse to match inputs for training NN\n",
    "#     features_list = rev_features.tolist()\n",
    "# #     print(cnt)\n",
    "#     append_vector(memory, features_list)\n",
    "#     # we need to fill out the buffer else it's filled with zeros\n",
    "#     if cnt >= MEMORY_LEN:\n",
    "#         X_vec = np.array(memory).reshape(1, MEMORY_LEN*FEATURE_SIZE)\n",
    "# #         print(X_vec)\n",
    "# #         t_temp = time.time()\n",
    "#         y_predict = model.predict_proba(X_vec)[:, 1][0]\n",
    "# #         print(time.time()-t_temp)\n",
    "#         temp_overlap = (y_predict >= threshold)\n",
    "        \n",
    "#         print('prediction', y_predict)\n",
    "#         print('Model detected overlap', temp_overlap)\n",
    "#         print('Geometric overlap', circle_check(df.x.values[cnt],df.y.values[cnt], robot=robot_flag))\n",
    "#         print('Time', df.t.values[cnt])\n",
    "#         print()\n",
    "# #         print(pred_memory)\n",
    "#         append_vector(pred_memory, temp_overlap)\n",
    "#         if all(pred_memory):\n",
    "#             print('SWITCH!!!!')\n",
    "#             switch = True\n",
    "            \n",
    "#             cnt = len(df)-1\n",
    "#     cnt += 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "ceebc280",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "5d31f8c8",
   "metadata": {},
   "outputs": [],
   "source": [
    "start_idx = df.index[df['Case'].diff() == 1].tolist()\n",
    "print(start_idx)\n",
    "print(df.loc[start_idx[0]].t)\n",
    "print(df.loc[start_idx[1]].t)\n",
    "# print(df.loc[start_idx[2]].t)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "6d84ab0b",
   "metadata": {
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "plt.figure()\n",
    "plt.scatter(df.t.values, df.Case.values)\n",
    "plt.grid()\n",
    "plt.title('Case')\n",
    "plt.show()\n",
    "\n",
    "plt.figure()\n",
    "plt.plot(df.t.values, df.Fx.values)\n",
    "plt.grid()\n",
    "plt.title('Fx')\n",
    "plt.show()\n",
    "\n",
    "plt.figure()\n",
    "plt.plot(df.t.values, df.Fy.values)\n",
    "plt.grid()\n",
    "plt.title('Fy')\n",
    "plt.show()\n",
    "\n",
    "plt.figure()\n",
    "plt.plot(df.t.values, df.Fz.values)\n",
    "plt.grid()\n",
    "plt.title('Fz')\n",
    "plt.show()\n",
    "\n",
    "\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "063f578d",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "2ba343db",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "60ad53ef",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "6311f2bc",
   "metadata": {},
   "outputs": [],
   "source": [
    "temp_vec = np.where(df.Case == 1)\n",
    "for x in temp_vec:\n",
    "    print(x)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "34362750",
   "metadata": {},
   "outputs": [],
   "source": [
    "df.loc[17585].t-df.loc[17229].t"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "c420d2dd",
   "metadata": {},
   "outputs": [],
   "source": [
    "a = pd.DataFrame([0,0,0,0,0,1,1,1,0,0,0,1,1,0,0],columns=['Case'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "976baf00",
   "metadata": {},
   "outputs": [],
   "source": [
    "a"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "36ba41bf",
   "metadata": {},
   "outputs": [],
   "source": [
    "start_idx = a.index[a['Case'].diff() == 1].tolist()\n",
    "stop_idx = a.index[a['Case'].diff() == -1].tolist()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "0a66f1fd",
   "metadata": {},
   "outputs": [],
   "source": [
    "start_idx"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "ef858dec",
   "metadata": {},
   "outputs": [],
   "source": [
    "stop_idx"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "08ad20a7",
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "a.loc[0:0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "d8dea9ce",
   "metadata": {},
   "outputs": [],
   "source": [
    "new_df = a.iloc[start_idx[0]:stop_idx[0]]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "1f685a70",
   "metadata": {},
   "outputs": [],
   "source": [
    "new_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "5b23f330",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "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": "3.8.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}