cartpole_obs_path_visualization_mpc.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "obs_idx: \n",
      "3\n",
      "p_idx:\n",
      "1829\n",
      "generated.\n",
      "(7, 2)\n"
     ]
    }
   ],
   "source": [
    "## from ctypes import *\n",
    "#ctypes.cdll.LoadLibrary('')\n",
    "#lib1 = CDLL(\"deps/sparse_rrt/deps/trajopt/build/lib/libsco.so\")\n",
    "#lib2 = CDLL(\"deps/sparse_rrt/deps/trajopt/build/lib/libutils.so\")\n",
    "\n",
    "import sys\n",
    "sys.path.append('/media/arclabdl1/HD1/YLmiao/mpc-mpnet-cuda-yinglong/deps/sparse_rrt-1')\n",
    "sys.path.append('.')\n",
    "\n",
    "from sparse_rrt.planners import SST\n",
    "#from env.cartpole_obs import CartPoleObs\n",
    "#from env.cartpole import CartPole\n",
    "#from sparse_rrt.systems import standard_cpp_systems\n",
    "from sparse_rrt import _sst_module\n",
    "import numpy as np\n",
    "import time\n",
    "from plan_utility.line_line_cc import *\n",
    "import pickle\n",
    "from sparse_rrt import _deep_smp_module, _sst_module\n",
    "import os\n",
    "obs_list = []\n",
    "LENGTH = 20.\n",
    "width = 6.\n",
    "near = width * 1.2\n",
    "# convert from obs to point cloud\n",
    "# load generated point cloud\n",
    "obs_list_total = []\n",
    "obc_list_total = []\n",
    "for i in range(10):\n",
    "    file = open('/media/arclabdl1/HD1/YLmiao/data/kinodynamic/cartpole_obs/obs_%d.pkl' % (i), 'rb')\n",
    "    obs_list_total.append(pickle.load(file))\n",
    "    file = open('/media/arclabdl1/HD1/YLmiao/data/kinodynamic/cartpole_obs/obc_%d.pkl' % (i), 'rb')\n",
    "    obc_list_total.append(pickle.load(file))\n",
    "\n",
    "#[(0, 932), (1, 935), (2, 923), (8, 141), (5,931), (7, 927)]\n",
    "# (5,931), (6, 286)\n",
    "\n",
    "obs_idx = 3\n",
    "\n",
    "p_idx = 1829\n",
    "\n",
    "# ignore_list = [(8,1914), (0, 1810), (1, 1803), (2, 1809), (3, 1829), (4, 1814), (5, 1826), (7, 1886)]\n",
    "# found = False\n",
    "# print(os.path.isfile('/media/arclabdl1/HD1/YLmiao/mpc-mpnet-cuda-yinglong/results/cpp_full/cartpole_obs/default_small_model_batch/paths/path_%d_%d.npy' % (obs_idx, p_idx)))\n",
    "# for obs_idx in range(6,10):\n",
    "#     for p_idx in range(1800,2000):\n",
    "#         if (obs_idx, p_idx) in ignore_list:\n",
    "#             continue\n",
    "#         if not os.path.isfile('/media/arclabdl1/HD1/YLmiao/mpc-mpnet-cuda-yinglong/results/cpp_full/cartpole_obs/default_small_model_batch/paths/path_%d_%d.npy' % (obs_idx, p_idx)):\n",
    "#             continue\n",
    "#         cost = np.load('/media/arclabdl1/HD1/YLmiao/mpc-mpnet-cuda-yinglong/results/cpp_full/cartpole_obs/default_small_model_batch/paths/path_%d_%d.npy' % (obs_idx, p_idx),\n",
    "#                allow_pickle=True)[2].sum()\n",
    "#         data_cost = open('/media/arclabdl1/HD1/YLmiao/data/kinodynamic/cartpole_obs/%d/cost_%d.pkl' % (obs_idx, p_idx), 'rb')\n",
    "#         data_cost = pickle.load(data_cost).sum()\n",
    "#         if cost < 0.4*data_cost:\n",
    "#             found = True\n",
    "#             break\n",
    "#     if found:\n",
    "#         break\n",
    "\n",
    "print('obs_idx: ')\n",
    "print(obs_idx)\n",
    "print('p_idx:')\n",
    "print(p_idx)\n",
    "        \n",
    "\n",
    "# Create custom system\n",
    "#obs_list = [[-10., -3.],\n",
    "#            [0., 3.],\n",
    "#            [10, -3.]]\n",
    "obs_list = obs_list_total[obs_idx]\n",
    "obc_list = obc_list_total[obs_idx]\n",
    "print('generated.')\n",
    "print(obs_list.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "\"\\nfor p_idx in range(900,1000):\\n    sgs = open('/media/arclabdl1/HD1/YLmiao/data/kinodynamic/acrobot_obs_backup/%d/start_goal_%d.pkl' % (obs_idx, p_idx), 'rb')\\n    sgs = pickle.load(sgs)\\n    dis = np.abs(sgs[1]-np.array([np.pi, 0, 0., 0.]))\\n    if dis[0] >= np.pi:\\n        dis[0] = 2*np.pi - dis[0]\\n    if dis[1] >= np.pi:\\n        dis[1] = 2*np.pi - dis[1]\\n        \\n    if np.linalg.norm(dis) <= 1e-1:\\n        break\\nprint(p_idx)\\n\""
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# search for goal\n",
    "\"\"\"\n",
    "for p_idx in range(900,1000):\n",
    "    sgs = open('/media/arclabdl1/HD1/YLmiao/data/kinodynamic/acrobot_obs_backup/%d/start_goal_%d.pkl' % (obs_idx, p_idx), 'rb')\n",
    "    sgs = pickle.load(sgs)\n",
    "    dis = np.abs(sgs[1]-np.array([np.pi, 0, 0., 0.]))\n",
    "    if dis[0] >= np.pi:\n",
    "        dis[0] = 2*np.pi - dis[0]\n",
    "    if dis[1] >= np.pi:\n",
    "        dis[1] = 2*np.pi - dis[1]\n",
    "        \n",
    "    if np.linalg.norm(dis) <= 1e-1:\n",
    "        break\n",
    "print(p_idx)\n",
    "\"\"\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "\n",
    "params = {\n",
    "        'n_problem': 1,\n",
    "        'n_sample': 1024,\n",
    "        'n_elite': 128,\n",
    "        'n_t': 1,\n",
    "        'max_it': 100,\n",
    "        #'converge_r': 5e-1,\n",
    "        'converge_r': 1e-5,\n",
    "    \n",
    "        'dt': 2e-3,\n",
    "        'mu_u': np.array([0.]),\n",
    "        'sigma_u': np.array([400]),\n",
    "        'mu_t': 4e-1,\n",
    "        'sigma_t': 4e-1,\n",
    "        't_max': 0.6,\n",
    "\n",
    "        'verbose': True,# False,#\n",
    "        'step_size': 0,\n",
    "\n",
    "        \"goal_radius\": 1.5,\n",
    "\n",
    "        \"sst_delta_near\": 1,\n",
    "        \"sst_delta_drain\": 1e-1,\n",
    "        \"goal_bias\": 0.0,\n",
    "\n",
    "        \"width\": 4,\n",
    "        \"hybrid\": False,#True,# \n",
    "        \"hybrid_p\": 0.0,\n",
    "        \n",
    "        \"min_time_steps\": 5,\n",
    "        \"max_time_steps\": 100,\n",
    "    \n",
    "        \"cost_samples\": 1,\n",
    "        \"mpnet_weight_path\":\"/home/arclabdl1/YLmiao/kinodynamics/mpc-mpnet-cuda-yinglong/mpnet/exported/output/cartpole_obs/mpnet_10k_external_small_model.pt\",\n",
    "        \"cost_predictor_weight_path\": \"/home/arclabdl1/YLmiao/kinodynamics/mpc-mpnet-cuda-yinglong/mpnet/exported/output/cartpole_obs/cost_10k.pt\",\n",
    "        \"cost_to_go_predictor_weight_path\": \"/home/arclabdl1/YLmiao/kinodynamics/mpc-mpnet-cuda-yinglong/mpnet/exported/output/cartpole_obs/cost_to_go_10k.pt\",\n",
    "\n",
    "        \"refine\": False,\n",
    "        \"using_one_step_cost\": False,\n",
    "        \"refine_lr\": 0,\n",
    "        \"refine_threshold\": 0,\n",
    "        \"device_id\": \"cuda:3\",\n",
    "\n",
    "        \"cost_reselection\": False,\n",
    "        \"number_of_iterations\": 2000,\n",
    "        \"weights_array\": [1, .5, 1, .5],\n",
    "\n",
    "    }"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[array([21.61279251,  0.        ,  1.52666725,  0.        ]), array([-23.15280384, -34.14387832,  -2.43323525,   0.43542129])]\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "'\\npath = np.array(\\n[[ 0.,          0.,          0.,          0.        ],\\n [ 0.1708568,  -0.39391056,  0.19467657, -0.62469995],\\n [ 0.28792115, -0.91607445,  0.44898465, -1.63658985],\\n [-0.95989468,  1.96289103,  0.9317956,   1.43428469],\\n [-0.30129242,  2.94121546, -1.83117369,  0.90737314],\\n [-0.70691949, -3.12132232, -1.05535802,  0.6535526 ],\\n [-0.82442304, -3.03540106,  0.01787491,  0.13220202],\\n [ 0.52775011,  2.27083079,  0.37915111, -0.44081038],\\n [-1.07891294, -2.37855552, -0.30601515,  1.50195129],\\n [ 1.64625095,  2.19734582,  1.28417728, -1.54819197],\\n [ 1.75929464,  2.02474293,  0.19036604, -0.80510894],\\n [-1.9036383,  -1.9494513,  -1.02537086,  0.82188096],\\n [ 2.3872588,   1.49882542,  1.73340551, -2.18197651],])\\n'"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# load path\n",
    "path, control, time = np.load('/media/arclabdl1/HD1/YLmiao/mpc-mpnet-cuda-yinglong/results/cpp_full/cartpole_obs/default_small_model_batch/paths/path_%d_%d.npy' % (obs_idx, p_idx),\n",
    "               allow_pickle=True)\n",
    "\n",
    "sgs = open('/media/arclabdl1/HD1/YLmiao/data/kinodynamic/cartpole_obs/%d/start_goal_%d.pkl' % (obs_idx, p_idx), 'rb')\n",
    "sgs = pickle.load(sgs)\n",
    "print(sgs)\n",
    "\"\"\"\n",
    "# (7, 927)\n",
    "path = np.array(\n",
    "[[ 0.,          0.,          0.,          0.        ],\n",
    " [-0.30235064,  0.66357245, -1.12404058,  2.63216073],\n",
    " [-0.42513114,  0.98700913, -0.29263858,  1.22161321],\n",
    " [ 0.81097696, -2.21064598, -1.87366238, -0.44346807],\n",
    " [ 0.43625224, -2.17189627, -2.81821699,  1.07151433],\n",
    " [ 1.30897045,  2.50860498,  0.25455659, -1.13373327],\n",
    " [ 1.30296453,  2.41413713, -0.5262052, -0.66861023],\n",
    " [-1.58742542, -1.43178197, -2.36883756,  4.15048263],\n",
    " [-1.88335465, -0.77853767, -0.38750337,  2.06091217],\n",
    " [ 1.86185939, -1.89328253,  1.66468725,  3.40569328],\n",
    " [ 1.88586336, -0.82441263, -1.47474734,  5.44456514],\n",
    " [ 3.02422565,  2.29042714, -1.37086011,  0.0215158 ],\n",
    " [ 2.59949331,  2.31621808, -1.51831319,  0.3535678 ],\n",
    " [ 2.42948197,  2.37665629, -2.00484496,  0.99423899],\n",
    " [-1.57991341, -1.36811145, -2.2772159,   2.70299448],\n",
    " [ 2.80394195,  0.71981286,  1.12148533, -1.98584426],\n",
    " [ 3.00295891,  0.34951812,  0.73202747, -1.4486524 ],\n",
    " [ 3.13364725,  0.10208493, -0.41396637,  0.97824909]])\n",
    "\n",
    "# (1, 916)\n",
    "path = np.array(\n",
    " [[ 0.,          0.,          0.,          0.        ],\n",
    " [-0.14637182,  0.36741907,  0.16802393, -0.09198922],\n",
    " [ 0.33910537, -0.53233128,  1.01860863, -2.2462016 ],\n",
    " [ 0.45433237, -0.84326897,  0.00979799, -0.67701909],\n",
    " [ 0.15750956, -0.76430055, -1.76589893,  2.27511201],\n",
    " [ 0.05599173,  2.06468595,  2.89323376, -1.13509309],\n",
    " [ 0.70446343,  1.75184533,  1.34005915, -0.63608721],\n",
    " [-0.6347062,  -3.02094395, -3.14349578,  2.88615537],\n",
    " [-1.29986757, -2.26094529, -0.70598324,  1.6203048 ],\n",
    " [ 1.52734201,  1.99942129,  2.90282734, -3.52070003],\n",
    " [ 1.89659985,  1.48138215,  1.52561572, -2.66617007],\n",
    " [ 2.02352651,  1.23006466,  0.90312322, -2.28778868],\n",
    " [ 2.09466351,  0.73102972, -0.39847056, -1.59188   ],\n",
    " [ 2.07385742,  0.67449809, -0.88810876, -0.87647985]])\n",
    "\"\"\"\n",
    "# (8, 914)\n",
    "\"\"\"\n",
    "path = np.array(\n",
    "[[ 0.,          0.,          0.,          0.        ],\n",
    " [ 0.1708568,  -0.39391056,  0.19467657, -0.62469995],\n",
    " [ 0.28792115, -0.91607445,  0.44898465, -1.63658985],\n",
    " [-0.95989468,  1.96289103,  0.9317956,   1.43428469],\n",
    " [-0.30129242,  2.94121546, -1.83117369,  0.90737314],\n",
    " [-0.70691949, -3.12132232, -1.05535802,  0.6535526 ],\n",
    " [-0.82442304, -3.03540106,  0.01787491,  0.13220202],\n",
    " [ 0.52775011,  2.27083079,  0.37915111, -0.44081038],\n",
    " [-1.07891294, -2.37855552, -0.30601515,  1.50195129],\n",
    " [ 1.64625095,  2.19734582,  1.28417728, -1.54819197],\n",
    " [ 1.75929464,  2.02474293,  0.19036604, -0.80510894],\n",
    " [-1.9036383,  -1.9494513,  -1.02537086,  0.82188096],\n",
    " [ 2.3872588,   1.49882542,  1.73340551, -2.18197651],])\n",
    "\"\"\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "planner = _deep_smp_module.DSSTMPCWrapper(\n",
    "    system_type='cartpole_obs',\n",
    "    solver_type=\"cem\",\n",
    "    start_state=np.array(path[0]),\n",
    "#             goal_state=np.array(ref_path[-1]),\n",
    "    goal_state=np.array(sgs[-1]),\n",
    "    goal_radius=params['goal_radius'],\n",
    "    random_seed=0,\n",
    "    sst_delta_near=params['sst_delta_near'],\n",
    "    sst_delta_drain=params['sst_delta_drain'],\n",
    "    obs_list=obs_list,\n",
    "    width=params['width'],\n",
    "    verbose=params['verbose'],\n",
    "    mpnet_weight_path=params['mpnet_weight_path'], \n",
    "    cost_predictor_weight_path=params['cost_predictor_weight_path'],\n",
    "    cost_to_go_predictor_weight_path=params['cost_to_go_predictor_weight_path'],\n",
    "    num_sample=params['cost_samples'],\n",
    "    np=params['n_problem'], ns=params['n_sample'], nt=params['n_t'], ne=params['n_elite'], max_it=params['max_it'],\n",
    "    converge_r=params['converge_r'], mu_u=params['mu_u'], std_u=params['sigma_u'], mu_t=params['mu_t'], \n",
    "    std_t=params['sigma_t'], t_max=params['t_max'], step_size=params['step_size'], integration_step=params['dt'], \n",
    "    device_id=params['device_id'], refine_lr=params['refine_lr'],\n",
    "    weights_array=params['weights_array'],\n",
    "    obs_voxel_array=obc_list.reshape(-1)\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from plan_utility.line_line_cc import line_line_cc\n",
    "\n",
    "\n",
    "def enforce_bounds(state):\n",
    "    '''\n",
    "\n",
    "    check if state satisfies the bound\n",
    "    apply threshold to velocity and angle\n",
    "    return a new state toward which the bound has been enforced\n",
    "    '''\n",
    "    I = 10\n",
    "    L = 2.5\n",
    "    M = 10\n",
    "    m = 5\n",
    "    g = 9.8\n",
    "    H = 0.5  # cart\n",
    "    # define the name for each state index and action index\n",
    "    STATE_X, STATE_V, STATE_THETA, STATE_W = 0, 1, 2, 3\n",
    "    CONTROL_A = 0\n",
    "    # define boundary\n",
    "    MIN_X = -30\n",
    "    MAX_X = 30\n",
    "    MIN_V = -40\n",
    "    MAX_V = 40\n",
    "    MIN_W = -2\n",
    "    MAX_W = 2\n",
    "    # obstacle information\n",
    "    OBS_W = 4\n",
    "    new_state = np.array(state)\n",
    "    \"\"\"\n",
    "    if state[STATE_V] < MIN_V/30.:\n",
    "        new_state[STATE_V] = MIN_V/30.\n",
    "    elif state[STATE_V] > MAX_V/30.:\n",
    "        new_state[STATE_V] = MAX_V/30.\n",
    "    \"\"\"\n",
    "\n",
    "    if state[STATE_V] < MIN_V:\n",
    "        new_state[STATE_V] = MIN_V\n",
    "    elif state[STATE_V] > MAX_V:\n",
    "        new_state[STATE_V] = MAX_V\n",
    "\n",
    "    if state[STATE_THETA] < -np.pi:\n",
    "        new_state[STATE_THETA] += 2*np.pi\n",
    "    elif state[STATE_THETA] > np.pi:\n",
    "        new_state[STATE_THETA] -= 2*np.pi\n",
    "\n",
    "    if state[STATE_W] < MIN_W:\n",
    "        new_state[STATE_W] = MIN_W\n",
    "    elif state[STATE_W] > MAX_W:\n",
    "        new_state[STATE_W] = MAX_W\n",
    "    return new_state\n",
    "\n",
    "\n",
    "def IsInCollision(x, obc, obc_width=4.):\n",
    "    I = 10\n",
    "    L = 2.5\n",
    "    M = 10\n",
    "    m = 5\n",
    "    g = 9.8\n",
    "    H = 0.5\n",
    "\n",
    "    STATE_X = 0\n",
    "    STATE_V = 1\n",
    "    STATE_THETA = 2\n",
    "    STATE_W = 3\n",
    "    CONTROL_A = 0\n",
    "\n",
    "    MIN_X = -30\n",
    "    MAX_X = 30\n",
    "    MIN_V = -40\n",
    "    MAX_V = 40\n",
    "    MIN_W = -2\n",
    "    MAX_W = 2\n",
    "\n",
    "\n",
    "    if x[0] < MIN_X or x[0] > MAX_X:\n",
    "        return True\n",
    "\n",
    "    H = 0.5\n",
    "    pole_x1 = x[0]\n",
    "    pole_y1 = H\n",
    "    pole_x2 = x[0] + L * np.sin(x[2])\n",
    "    pole_y2 = H + L * np.cos(x[2])\n",
    "\n",
    "\n",
    "    for i in range(len(obc)):\n",
    "        for j in range(0, 8, 2):\n",
    "            x1 = obc[i][j]\n",
    "            y1 = obc[i][j+1]\n",
    "            x2 = obc[i][(j+2) % 8]\n",
    "            y2 = obc[i][(j+3) % 8]\n",
    "            if line_line_cc(pole_x1, pole_y1, pole_x2, pole_y2, x1, y1, x2, y2):\n",
    "                return True\n",
    "    return False\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def wrap_angle(x, system):\n",
    "    circular = system.is_circular_topology()\n",
    "    res = np.array(x)\n",
    "    for i in range(len(x)):\n",
    "        if circular[i]:\n",
    "            # use our previously saved version\n",
    "            res[i] = x[i] - np.floor(x[i] / (2*np.pi))*(2*np.pi)\n",
    "            if res[i] > np.pi:\n",
    "                res[i] = res[i] - 2*np.pi\n",
    "    return res\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from visual.visualizer import Visualizer\n",
    "import matplotlib.pyplot as plt\n",
    "import matplotlib.animation as animation\n",
    "import matplotlib as mpl\n",
    "import matplotlib.patches as patches\n",
    "from matplotlib.colors import LinearSegmentedColormap\n",
    "class CartpoleVisualizer(Visualizer):\n",
    "    def __init__(self, system, params):\n",
    "        super(CartpoleVisualizer, self).__init__(system, params)\n",
    "        self.dt = 2\n",
    "        self.fig1 = plt.figure(figsize=(25,6))\n",
    "        self.fig2 = plt.figure(figsize=(25,20))\n",
    "        self.ax1 = self.fig1.add_subplot(1,1,1)\n",
    "        self.ax2 = self.fig2.add_subplot(1,1,1)\n",
    "\n",
    "    def _init(self):\n",
    "        ##### handle the animation\n",
    "        # clear the current ax\n",
    "        print(\"in init\")\n",
    "        ax = self.ax1\n",
    "        ax.clear()\n",
    "        # add patches\n",
    "        state = self.states[0]\n",
    "        self.pole = patches.Rectangle((state[0]-self.params['pole_w']/2,self.params['cart_h']),\\\n",
    "                                       self.params['pole_w'],self.params['pole_l'],\\\n",
    "                                      linewidth=.5,edgecolor='green',facecolor='green')\n",
    "        self.cart = patches.Rectangle((state[0]-self.params['cart_w']/2,0),\\\n",
    "                                       self.params['cart_w'],self.params['cart_h'],\\\n",
    "                                      linewidth=.5,edgecolor='green',facecolor='green')\n",
    "        self.recs = []\n",
    "        self.recs.append(self.pole)\n",
    "        self.recs.append(self.cart)\n",
    "        for i in range(len(self.obs)):\n",
    "            x, y = self.obs[i]\n",
    "            obs = patches.Rectangle((x-self.params['obs_w']/2,y-params['obs_h']/2),\\\n",
    "                                       self.params['obs_w'],self.params['obs_h'],\\\n",
    "                                      linewidth=.5,edgecolor='black',facecolor='black')\n",
    "            self.recs.append(obs)\n",
    "            ax.add_patch(obs)\n",
    "        # transform pole according to state\n",
    "        t = mpl.transforms.Affine2D().rotate_deg_around(state[0], self.params['cart_h'], \\\n",
    "                                                        -state[2]/np.pi * 180) + ax.transData\n",
    "        self.pole.set_transform(t)\n",
    "        ax.add_patch(self.pole)\n",
    "        ax.add_patch(self.cart)\n",
    "\n",
    "        #### handle search space\n",
    "        ax = self.ax2\n",
    "        ax.clear()\n",
    "        ax.set_xlim(-25, 25)\n",
    "        ax.set_ylim(-np.pi, np.pi)\n",
    "\n",
    "        dx = 1\n",
    "        dtheta = 0.1\n",
    "        feasible_points = []\n",
    "        infeasible_points = []\n",
    "        imin = 0\n",
    "        imax = int(2*30./dx)\n",
    "        jmin = 0\n",
    "        jmax = int(2*np.pi/dtheta)\n",
    "\n",
    "        for i in range(imin, imax):\n",
    "            for j in range(jmin, jmax):\n",
    "                x = np.array([dx*i-30, 0., dtheta*j-np.pi, 0.])\n",
    "                if IsInCollision(x, self.cc_obs):\n",
    "                    infeasible_points.append(x)\n",
    "                else:\n",
    "                    feasible_points.append(x)\n",
    "        feasible_points = np.array(feasible_points)\n",
    "        infeasible_points = np.array(infeasible_points)\n",
    "        \n",
    "        print('feasible points')\n",
    "        print(feasible_points)\n",
    "        print('infeasible points')\n",
    "        print(infeasible_points)\n",
    "        #scat_feas =ax.scatter(feasible_points[:,0], feasible_points[:,2], c='yellow')\n",
    "        scat_infeas = ax.scatter(infeasible_points[:,0], infeasible_points[:,2], c='black')\n",
    "\n",
    "        #self.recs.append(scat_feas)\n",
    "        self.recs.append(scat_infeas)\n",
    "\n",
    "        scat_state = ax.scatter(state[0], state[2], c='green')\n",
    "        self.recs.append(scat_state)\n",
    "        print(\"after init\")\n",
    "\n",
    "        return self.recs\n",
    "    def _animate(self, i):\n",
    "        print('animating, frame %d/%d' % (i, self.total))\n",
    "        \n",
    "        ax = self.ax1\n",
    "        ax.set_xlim(-25, 25)\n",
    "        ax.set_ylim(-6, 6)\n",
    "        state = self.states[i]\n",
    "        self.recs[0].set_xy((state[0]-self.params['pole_w']/2,self.params['cart_h']))\n",
    "        t = mpl.transforms.Affine2D().rotate_deg_around(state[0], self.params['cart_h'], \\\n",
    "                                                        -state[2]/np.pi * 180) + ax.transData\n",
    "        self.recs[0].set_transform(t)\n",
    "        self.recs[1].set_xy((state[0]-self.params['cart_w']/2,0))\n",
    "\n",
    "\n",
    "        # handle search space\n",
    "        ax = self.ax2\n",
    "        ax.set_xlim(-25, 25)\n",
    "        ax.set_ylim(-np.pi, np.pi)\n",
    "        self.recs[-1].set_offsets([state[0], state[2]])\n",
    "        # print location of cart\n",
    "        return self.recs\n",
    "\n",
    "\n",
    "    def animate(self, states, actions, costs, obstacles, sg, wrap_system):\n",
    "        '''\n",
    "        given a list of states, actions and obstacles, animate the robot\n",
    "        '''\n",
    "\n",
    "        new_obs_i = []\n",
    "        obs_width = 4.0\n",
    "        for k in range(len(obstacles)):\n",
    "            obs_pt = []\n",
    "            obs_pt.append(obstacles[k][0]-obs_width/2)\n",
    "            obs_pt.append(obstacles[k][1]-obs_width/2)\n",
    "            obs_pt.append(obstacles[k][0]-obs_width/2)\n",
    "            obs_pt.append(obstacles[k][1]+obs_width/2)\n",
    "            obs_pt.append(obstacles[k][0]+obs_width/2)\n",
    "            obs_pt.append(obstacles[k][1]+obs_width/2)\n",
    "            obs_pt.append(obstacles[k][0]+obs_width/2)\n",
    "            obs_pt.append(obstacles[k][1]-obs_width/2)\n",
    "            new_obs_i.append(obs_pt)\n",
    "        obs_i = new_obs_i\n",
    "        self.cc_obs = obs_i\n",
    "\n",
    "        # transform the waypoint states and actions into trajectory\n",
    "        traj = []\n",
    "        s = states[0]\n",
    "        for i in range(len(states)-1):\n",
    "            s = states[i]\n",
    "            print('state: %d, remaining: %d' % (i, len(states)-i))\n",
    "            #action = actions[i]\n",
    "            \n",
    "            # connect from this state to next\n",
    "            solution_u, solution_t = planner.steer_solution(states[i], states[i+1])\n",
    "            print('solution_u:')\n",
    "            print(solution_u)\n",
    "            print('solution_t:')\n",
    "            print(solution_t)\n",
    "            for j in range(len(solution_u)):\n",
    "                action = solution_u[j]\n",
    "                num_steps = int(np.round(solution_t[j]/self.params['integration_step']))\n",
    "\n",
    "                for k in range(num_steps):\n",
    "                    traj.append(np.array(s))\n",
    "                    #print(\"porpagating...\")\n",
    "                    #print(s)\n",
    "                    #print('st:')\n",
    "                    #print(sT)\n",
    "                    s = self.system(s, action, self.params['integration_step'])\n",
    "                    s = enforce_bounds(s)\n",
    "                    assert not IsInCollision(s, obs_i)\n",
    "            print('after propagation, state: ', s)\n",
    "            print('state to reach: ', states[i+1])\n",
    "            \n",
    "        traj = np.array(traj)\n",
    "        print(\"animating...\")\n",
    "        # animate\n",
    "        self.states = traj\n",
    "        self.obs = obstacles\n",
    "        print(len(self.states))\n",
    "        self.total = len(self.states)\n",
    "        self._init()\n",
    "        \n",
    "        to_plot_list_x = []\n",
    "        to_plot_list_y = []\n",
    "        traj_recs = []\n",
    "        plot_step_sz = 50\n",
    "        for i in list(range(1,len(traj))) + [0]:\n",
    "            if i % plot_step_sz == 0:\n",
    "                # plot the scene change\n",
    "                if i == 0:\n",
    "                    pole_color = \"green\"\n",
    "                    cart_color = \"green\"\n",
    "                elif i+plot_step_sz >= len(traj):\n",
    "                    pole_color = \"red\"\n",
    "                    cart_color = \"red\"\n",
    "                else:\n",
    "                    pole_color = \"blue\"\n",
    "                    cart_color = \"blue\"\n",
    "                state = traj[i]\n",
    "                pole = patches.Rectangle((state[0]-self.params['pole_w']/2,self.params['cart_h']),\\\n",
    "                                               self.params['pole_w'],self.params['pole_l'],\\\n",
    "                                              linewidth=.5,edgecolor=pole_color,facecolor=pole_color)\n",
    "                cart = patches.Rectangle((state[0]-self.params['cart_w']/2,0),\\\n",
    "                                               self.params['cart_w'],self.params['cart_h'],\\\n",
    "                                              linewidth=.5,edgecolor=cart_color,facecolor=cart_color)\n",
    "\n",
    "                traj_recs.append(pole)\n",
    "                traj_recs.append(cart)\n",
    "                ax = self.ax1\n",
    "                ax.set_xlim(-25, 25)\n",
    "                ax.set_ylim(-6, 6)\n",
    "                ax.add_patch(pole)\n",
    "                ax.add_patch(cart)\n",
    "                pole.set_xy((state[0]-self.params['pole_w']/2,self.params['cart_h']))\n",
    "                t = mpl.transforms.Affine2D().rotate_deg_around(state[0], self.params['cart_h'], \\\n",
    "                                                                -state[2]/np.pi * 180) + ax.transData\n",
    "                pole.set_transform(t)\n",
    "                cart.set_xy((state[0]-self.params['cart_w']/2,0))\n",
    "                ax.axis('off')\n",
    "\n",
    "            # plot in the state space    \n",
    "            ax = self.ax2\n",
    "            traj_to_plot = wrap_angle(traj[i], wrap_system)\n",
    "            scat_state = ax.scatter(traj_to_plot[0], traj_to_plot[2], c='blue', s=25.0)\n",
    "        colors = ['green', 'red']\n",
    "        \n",
    "        #ax = self.ax1\n",
    "        #cm = LinearSegmentedColormap.from_list(\"Custom\", colors, N=len(to_plot_list_x))\n",
    "        \"\"\"\n",
    "        for i in range(len(to_plot_list_x)):\n",
    "            #ax.plot(to_plot_list_x[i], to_plot_list_y[i], alpha=float(i)/len(to_plot_list_x), c=cm(float(i)/len(to_plot_list_x)))\n",
    "            if i == 0:\n",
    "                ax.plot(to_plot_list_x[i], to_plot_list_y[i], alpha=1, c='green')\n",
    "            else:\n",
    "                ax.plot(to_plot_list_x[i], to_plot_list_y[i], alpha=float(i)/len(to_plot_list_x), c='blue')\n",
    "        \"\"\"\n",
    "            \n",
    "        # start and goal\n",
    "        ax = self.ax2\n",
    "        scat_state = ax.scatter(sg[0][0], sg[0][2], c='green', s=50.0)\n",
    "        scat_state = ax.scatter(sg[1][0], sg[1][2], c='red', s=50.0, marker='*')\n",
    "        plt.subplots_adjust(wspace=0, hspace=0)\n",
    "\n",
    "        ax.axis('off')\n",
    "\n",
    "        #plt.savefig(\"cartpole_mpnettree_%d_p_%d.png\" % (obs_idx, p_idx), bbox_inches='tight')\n",
    "        self.fig1.savefig(\"cartpole_mpnettree_%d_p_%d_workspace.pdf\" % (obs_idx, p_idx), bbox_inches='tight', pad_inches = 0)\n",
    "        self.fig2.savefig(\"cartpole_mpnettree_%d_p_%d_statespace.pdf\" % (obs_idx, p_idx), bbox_inches='tight', pad_inches = 0)\n",
    "        self.fig1.show()\n",
    "        self.fig2.show()\n",
    "        \n",
    "        \"\"\"\n",
    "        # for thesis\n",
    "        fig = plt.figure()\n",
    "        ax = plt.subplot(121)\n",
    "        ax.set_xlim(-40, 40)\n",
    "        ax.set_ylim(-40, 40)\n",
    "        fig.set_figheight(5)\n",
    "        fig.set_figwidth(10)\n",
    "        \n",
    "        for i in range(len(self.obs)):\n",
    "            x, y = self.obs[i]\n",
    "            obs = patches.Rectangle((x-self.params['obs_w']/2,y-params['obs_h']/2),\\\n",
    "                                       self.params['obs_w'],self.params['obs_h'],\\\n",
    "                                      linewidth=.5,edgecolor='black',facecolor='black')\n",
    "            self.recs.append(obs)\n",
    "            ax.add_patch(obs)\n",
    "        ax.plot(to_plot_list_x[0], to_plot_list_y[0], alpha=1, c='green')\n",
    "        ax.plot(to_plot_list_x[-1], to_plot_list_y[-1], alpha=1, c='red')\n",
    "        \n",
    "        ax = plt.subplot(122)\n",
    "        ax.set_xlim(-np.pi, np.pi)\n",
    "        ax.set_ylim(-np.pi, np.pi)\n",
    "        dtheta = 0.1\n",
    "        feasible_points = []\n",
    "        infeasible_points = []\n",
    "        imin = 0\n",
    "        imax = int(2*np.pi/dtheta)\n",
    "\n",
    "        for i in range(imin, imax):\n",
    "            for j in range(imin, imax):\n",
    "                x = np.array([dtheta*i-np.pi, dtheta*j-np.pi, 0., 0.])\n",
    "                if IsInCollision(x, self.cc_obs):\n",
    "                    infeasible_points.append(x)\n",
    "                else:\n",
    "                    feasible_points.append(x)\n",
    "        feasible_points = np.array(feasible_points)\n",
    "        infeasible_points = np.array(infeasible_points)\n",
    "\n",
    "        print('feasible points')\n",
    "        print(feasible_points)\n",
    "        print('infeasible points')\n",
    "        print(infeasible_points)\n",
    "        #scat_feas =ax.scatter(feasible_points[:,0], feasible_points[:,1], c='white')\n",
    "        scat_infeas = ax.scatter(infeasible_points[:,0], infeasible_points[:,1], c='black')\n",
    "        scat_state = ax.scatter(sg[0][0], sg[0][1], c='green', s=50.0)\n",
    "        scat_state = ax.scatter(sg[1][0], sg[1][1], c='red', s=50.0, marker='*')\n",
    "        \"\"\"\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "states:\n",
      "[[ 2.16127925e+01  0.00000000e+00  1.52666725e+00  0.00000000e+00]\n",
      " [ 1.81056202e+01 -1.16069728e+01  1.12304087e+00 -1.09157572e+00]\n",
      " [ 1.32454717e+01 -4.51228019e+00  2.16832720e-02 -2.00000000e+00]\n",
      " [ 1.27932889e+01  1.71642071e+00 -7.38316728e-01 -2.00000000e+00]\n",
      " [ 1.40054858e+01  6.45240631e+00 -1.31935471e+00 -1.86417600e+00]\n",
      " [ 1.41492520e+01 -5.68412176e+00 -1.91100238e+00 -5.30200926e-01]\n",
      " [ 1.06913346e+01 -1.19005826e+01 -2.05095649e+00 -2.29337817e-01]\n",
      " [ 1.20878967e+00 -1.97346203e+01 -2.06534509e+00  1.75615280e-01]\n",
      " [-1.42277323e+01 -3.17384573e+01 -1.97124933e+00  2.01978240e-01]\n",
      " [-1.85192684e+01 -3.43243052e+01 -1.94129059e+00  2.64290592e-01]\n",
      " [-2.17936903e+01 -3.38888816e+01 -1.90116705e+00  5.77435983e-01]]\n",
      "state: 0, remaining: 11\n",
      "solution_u:\n",
      "[[-300.]]\n",
      "solution_t:\n",
      "[0.6]\n",
      "after propagation, state:  [ 18.10559197 -11.60706846   1.12304177  -1.09157114]\n",
      "state to reach:  [ 18.10562019 -11.60697279   1.12304087  -1.09157572]\n",
      "state: 1, remaining: 10\n",
      "solution_u:\n",
      "[[113.56053229]]\n",
      "solution_t:\n",
      "[0.6]\n",
      "after propagation, state:  [13.2450457  -4.51375706  0.0216909  -2.        ]\n",
      "state to reach:  [13.24547166 -4.51228019  0.02168327 -2.        ]\n",
      "state: 2, remaining: 9\n",
      "solution_u:\n",
      "[[199.67550004]]\n",
      "solution_t:\n",
      "[0.44]\n",
      "after propagation, state:  [12.74527862  1.70839677 -0.85831528 -1.99941975]\n",
      "state to reach:  [12.79328889  1.71642071 -0.73831673 -2.        ]\n",
      "state: 3, remaining: 8\n",
      "solution_u:\n",
      "[[265.21741076]]\n",
      "solution_t:\n",
      "[0.316]\n",
      "after propagation, state:  [14.09438265  6.44879593 -1.35770074 -1.80343248]\n",
      "state to reach:  [14.00548576  6.45240631 -1.31935471 -1.864176  ]\n",
      "state: 4, remaining: 7\n",
      "solution_u:\n",
      "[[-300.]]\n",
      "solution_t:\n",
      "[0.58]\n",
      "after propagation, state:  [14.14925196 -5.68412176 -1.91100238 -0.53020093]\n",
      "state to reach:  [14.14925196 -5.68412176 -1.91100238 -0.53020093]\n",
      "state: 5, remaining: 6\n",
      "solution_u:\n",
      "[[-240.60859364]]\n",
      "solution_t:\n",
      "[0.392]\n",
      "after propagation, state:  [ 10.70599256 -11.91603481  -2.05163075  -0.2367741 ]\n",
      "state to reach:  [ 10.69133463 -11.90058263  -2.05095649  -0.22933782]\n",
      "state: 6, remaining: 5\n",
      "solution_u:\n",
      "[[-199.80092828]]\n",
      "solution_t:\n",
      "[0.6]\n",
      "after propagation, state:  [  1.20870148 -19.73491621  -2.06535996   0.17555815]\n",
      "state to reach:  [  1.20878967 -19.73462027  -2.06534509   0.17561528]\n",
      "state: 7, remaining: 4\n",
      "solution_u:\n",
      "[[-300.]]\n",
      "solution_t:\n",
      "[0.6]\n",
      "after propagation, state:  [-14.22773227 -31.73845726  -1.97124933   0.20197824]\n",
      "state to reach:  [-14.22773227 -31.73845726  -1.97124933   0.20197824]\n",
      "state: 8, remaining: 3\n",
      "solution_u:\n",
      "[[-300.]]\n",
      "solution_t:\n",
      "[0.128]\n",
      "after propagation, state:  [-18.45069927 -34.28455036  -1.94181684   0.26312314]\n",
      "state to reach:  [-18.51926837 -34.32430516  -1.94129059   0.26429059]\n",
      "state: 9, remaining: 2\n",
      "solution_u:\n",
      "[[138.00760304]]\n",
      "solution_t:\n",
      "[0.096]\n",
      "after propagation, state:  [-21.76831124 -33.35057748  -1.8984603    0.63373322]\n",
      "state to reach:  [-21.79369028 -33.88888164  -1.90116705   0.57743598]\n",
      "animating...\n",
      "2176\n",
      "in init\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/arclabdl1/YLmiao/kinodynamics/KinoMPNet/plan_utility/line_line_cc.py:2: RuntimeWarning: divide by zero encountered in double_scalars\n",
      "  uA = ((x4-x3)*(y1-y3) - (y4-y3)*(x1-x3)) / ((y4-y3)*(x2-x1) - (x4-x3)*(y2-y1))\n",
      "/home/arclabdl1/YLmiao/kinodynamics/KinoMPNet/plan_utility/line_line_cc.py:3: RuntimeWarning: divide by zero encountered in double_scalars\n",
      "  uB = ((x2-x1)*(y1-y3) - (y2-y1)*(x1-x3)) / ((y4-y3)*(x2-x1) - (x4-x3)*(y2-y1))\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "feasible points\n",
      "[[-30.           0.          -3.14159265   0.        ]\n",
      " [-30.           0.          -3.04159265   0.        ]\n",
      " [-30.           0.          -2.94159265   0.        ]\n",
      " ...\n",
      " [ 29.           0.           2.75840735   0.        ]\n",
      " [ 29.           0.           2.85840735   0.        ]\n",
      " [ 29.           0.           2.95840735   0.        ]]\n",
      "infeasible points\n",
      "[[-19.           0.           0.75840735   0.        ]\n",
      " [-19.           0.           0.85840735   0.        ]\n",
      " [-19.           0.           0.95840735   0.        ]\n",
      " ...\n",
      " [ 12.           0.          -1.04159265   0.        ]\n",
      " [ 12.           0.          -0.94159265   0.        ]\n",
      " [ 12.           0.          -0.84159265   0.        ]]\n",
      "after init\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1800x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1800x1440 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "params = {}\n",
    "params['pole_l'] = 2.5\n",
    "params['pole_w'] = 0.01\n",
    "params['cart_w'] = 1.\n",
    "params['cart_h'] = 0.5\n",
    "params['obs_w'] = 4\n",
    "params['obs_h'] = 4\n",
    "params['integration_step'] = 0.002\n",
    "system = _sst_module.CartPole()\n",
    "cpp_propagator = _sst_module.SystemPropagator()\n",
    "dynamics = lambda x, u, t: cpp_propagator.propagate(system, x, u, t)\n",
    "\n",
    "vis = CartpoleVisualizer(dynamics, params)\n",
    "states = path\n",
    "sgs[0] = wrap_angle(sgs[0], system)\n",
    "sgs[1] = wrap_angle(sgs[1], system)\n",
    "print('states:')\n",
    "print(states)\n",
    "anim = vis.animate(np.array(states), None, None, obs_list, np.array(sgs), system)\n",
    "#HTML(anim.to_html5_video())\n",
    "#anim.save('acrobot_env%d_path%d.mp4' % (obs_idx, p_idx))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "3\n",
      "1829\n"
     ]
    }
   ],
   "source": [
    "print(obs_idx)\n",
    "print(p_idx)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "anaconda-cloud": {},
  "kernelspec": {
   "display_name": "Python [conda env:ylmiao_mpc]",
   "language": "python",
   "name": "conda-env-ylmiao_mpc-py"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}