tensorflowでカルマンフィルタ

Tensorflowでカルマンフィルタを実装してみた．

• 計算グラフで1ループを実装
• session.runで呼び出し
• tensorflowの内部で変数を覚えておく
• Tips:
  • assignで内部変数に保存する
  • assignの返り値のoperatorをsess.runで呼び出す

numpyバージョン v.s. tensorflowバージョン．

コード

kalman.py

import numpy as np
import matplotlib.pyplot as plt
import time
import tensorflow as tf


def KalmanFilter_tf(T, x_obs, mu0, V0, A, b, Q, R):
    """
    tensorflowで実装したカルマンフィルタ．
    """

    x_predict = np.zeros((T, 2))
    x_filter = np.zeros((T, 2))
    V_predict = np.zeros((T, 2, 2))
    V_filter = np.zeros((T, 2, 2))

    x_predict[0] = mu0
    x_filter[0] = mu0
    V_predict[0] = V0
    V_filter[0] = V0

    g = tf.Graph()

    with g.as_default():  # 計算グラフ準備

        obs_ = tf.placeholder(tf.float32, name="obs", shape=(2, 1))

        A_ = tf.placeholder(tf.float32, name="A", shape=(2, 2))
        b_ = tf.placeholder(tf.float32, name="b", shape=(2, 1))
        Q_ = tf.placeholder(tf.float32, name="Q", shape=(2, 2))
        R_ = tf.placeholder(tf.float32, name="R", shape=(2, 2))

        mu0_ = tf.placeholder(tf.float32, name="mu0", shape=(2, 1))
        V0_ = tf.placeholder(tf.float32, name="V0", shape=(2, 2))

        mu = tf.Variable(tf.zeros((2, 1)), dtype=tf.float32, name="mu")
        V = tf.Variable(tf.zeros((2, 2)), dtype=tf.float32, name="V")

        mu0_init = tf.assign(mu, mu0_)
        V0_init = tf.assign(V, V0_)

        mu_ = A_ @ mu + b_
        V_ = A_ @ V @ tf.transpose(A_) + Q_

        S = V_ + R_
        K = V_ @ tf.matrix_inverse(S)

        # 内部変数で保存しておく
        mu_op = tf.assign(mu, mu_ + K @ (obs_ - mu_))
        V_op = tf.assign(V, V_ - K @ V_)

    with tf.Session(graph=g) as sess:

        # 初期化
        m, v = sess.run([mu0_init, V0_init],
                        feed_dict={mu0_: mu0.reshape((2, 1)), V0_: V0})

        start = time.time()

        for t in range(1, T):

            # 各時刻tでsess.runで呼び出す
            m, v = sess.run([mu_op, V_op],
                            feed_dict={A_: A, b_: b.reshape((2, 1)),
                                       Q_: Q, R_: R,
                                       obs_: x_obs[t].reshape((2, 1))
                                       })

            x_filter[t], V_filter[t] = m.transpose(), v

        elapsed_time = time.time() - start
        print("tensorflow: ", elapsed_time)

    return None, None, x_filter, V_filter


def KalmanFilter(T, x_obs, mu0, V0, A, b, Q, R):
    """
    こちらは普通のカルマンフィルタ．np実装．
    """

    mu = mu0
    V = V0

    x_predict = np.zeros((T, 2))
    x_filter = np.zeros((T, 2))
    V_predict = np.zeros((T, 2, 2))
    V_filter = np.zeros((T, 2, 2))

    x_predict[0] = mu.transpose()
    x_filter[0] = mu
    V_predict[0] = V
    V_filter[0] = V

    start = time.time()

    for t in range(1, T):

        mu_ = A @ mu + b
        V_ = A @ V @ A.transpose() + Q

        x_predict[t] = mu_
        V_predict[t] = V_

        S = V_ + R
        K = V_ @ np.linalg.inv(S)

        mu = mu_ + K @ (x_obs[t] - mu_)
        V = V_ - K @ V_

        x_filter[t] = mu
        V_filter[t] = V

    elapsed_time = time.time() - start
    print("numpy:      ", elapsed_time)

    return x_predict, V_predict, x_filter, V_filter


def main():

    # データ作成
    T = 20
    mu0 = np.array([100, 100])
    V0 = np.array([[10, 0],
                   [0, 10]])

    A = np.array([[1.001, 0.001],
                  [0, 0.99]])
    b = np.array([5, 10])
    Q = np.array([[20, 0],
                  [0, 20]])
    R = np.array([[20, 0],
                  [0, 20]])

    rvq = np.random.multivariate_normal(np.zeros(2), Q, T)
    rvr = np.random.multivariate_normal(np.zeros(2), R, T)
    obs = np.zeros((T, 2))
    obs[0] = mu0
    for i in range(1, T):
        obs[i] = A @ obs[i-1] + b + rvq[i] + rvr[i]
    # 作成終わり

    x_predict, V_predict, x_filter, V_filter = \
        KalmanFilter(T, obs, mu0, V0, A, b, Q, R)
    print(x_filter)

    x_predict_tf, V_predict_tf, x_filter_tf, V_filter_tf = \
        KalmanFilter_tf(T, obs, mu0, V0, A, b, Q, R)
    print(x_filter_tf)

    fig = plt.figure(figsize=(16, 9))
    ax = fig.gca()

    ax.scatter(obs[:, 0], obs[:, 1],
               s=10, alpha=1, marker="o", color='w', edgecolor='k')
    ax.plot(obs[:, 0], obs[:, 1],
            alpha=0.5, lw=1, color='k')

    ax.scatter(x_filter[:, 0], x_filter[:, 1],
               s=10, alpha=1, marker="o", color='r')
    ax.plot(x_filter[:, 0], x_filter[:, 1],
            alpha=0.5, lw=1, color='r')

    ax.scatter(x_filter_tf[:, 0], x_filter_tf[:, 1],
               s=10, alpha=1, marker="o", color='m')
    ax.plot(x_filter_tf[:, 0], x_filter_tf[:, 1],
            alpha=0.5, lw=1, color='m')

    plt.show()


if __name__ == "__main__":
    main()

結果

on Mac

当然，それほど速くない．（on macos, CPU tf 1.4.1, anaconda python 3.6）

numpy:       0.0004937648773193359
tensorflow:  0.00774383544921875

on Ubuntu

GPUならもっと遅い．(Ubuntu 16.04, GPU tf 1.6.0, Geforce 1080Ti, anaconda python 3.6）

numpy:       0.002942800521850586
(...tensorflow messages...)
tensorflow:  0.6563735008239746

(Ubuntu 16.04, CPU tf 1.6.0, anaconda python 3.6)

numpy:       0.002834796905517578
tensorflow:  0.027823209762573242