用Arduino Nano 33 BLE Sense实现投篮运动手柄设计 - Funpack8项目分享一

电子森林 2021-05-25 21:50

构建AI未来，Arm计算平台无处不在 如何增强能源基础设施的实时控制？

从Funpack8 Arduino上线到现在，快两个月过去了，也迎来了落幕时刻，但活动的落幕不代表板卡的积灰，大家后期还是可以使用这个板卡来研究机器学习。

而我们第一个分享的项目，作者使用TinyML完成设计，模拟制作了投篮运动手柄，其中还模拟机器学习通过声音控制不同颜色LED灯亮起，真切感受到机器学习带来的快乐。大家也通过参加Funpack第八期活动体会到Arduino的易用性和庞大的生态圈，简约而不简单。

项目链接：https://www.eetree.cn/project/detail/324

大家感兴趣的也可以联系我们工作人员，帮你们“搭桥牵线”，更进一步交流机器学习。接下来请欣赏aramy的投篮运动手柄项目制作。

项目介绍：Funpack第八期开发板Arduino Nano 33 BLE Sense。这是一款非常小巧的开发板，集成了非常多的传感器，尤其吸引我的是，TinyML在这个开发板上得到了支持。

这里我选择任务一：投篮运动手柄。利用NANO-33 BLE的加速度及角速度感应器，设计一款用于虚拟练习投篮的手柄。投篮者可通过手持手柄(开发板)模拟投篮动作，而手柄会根据初始物理参数(可由用户设定或者系统默认值)，对投篮者的命中率作出预判，并且可以将改进建议的信息反馈给投篮者。优秀的模拟投篮手柄的确可以帮助用户提高投篮命中率。作为作业去完成。

解决问题思路：这里绘制了个简单的投篮示意图（灵魂画手，请勿介意）。人手给篮球一个力，这个力持续一段时间，使篮球速度产生变化。然后篮球脱离手，仅仅受到重力影响，做自由落体运动。篮球到篮筐的水平距离、垂直距离都是已知的，那么求出篮球脱手时的速度，和速度方向即可。篮球脱手的速度可以由篮球受到的加速度和时间的积分获得，而这个开发板带有加速度传感器的，通过加速度传感器就可以测量出三轴的加速度，再乘以时间，就可以获得累加起来的速度了。至于说速度方向，自然是和加速度方向一致。用加速度和角速度可以算出开发板的四元数，拿到开发板在空间中的仰角就知道加速度与水平面的夹角了。有了夹角，计算水平加速度和垂直加速度。水平加速度可以计算水平方向的速度，垂直方向加速度，累加上重力加速度，就可以计算竖直方向的速度了。

#include . // this is for IMU on BLE 33#include#include "SensorFusion.h" //SFSF fusion;
float gx, gy, gz, ax, ay, az, mx, my, mz;float pitch, roll, yaw;float deltat;
void setup() {  Serial.begin(115200);  if (!IMU.begin()) {    Serial.println("Failed to initialize IMU!");    while (1);  }
  if (!BLE.begin()) {    Serial.println("starting BLE failed!");    while (1);  }  delay(3000);
}
void loop() {  if (IMU.accelerationAvailable() && IMU.gyroscopeAvailable() && IMU.magneticFieldAvailable()) {    IMU.readAcceleration(ax, ay, az);    IMU.readGyroscope(gx, gy, gz);    IMU.readMagneticField(mx, my, mz);    gx = gx - 1.77;     //校零    gy = gy + 0.521;    gz = gz + 0.431;    float gyroScale = 3.1415926f / 180;    gx = gx * gyroScale;    gy = gy * gyroScale;    gz = gz * gyroScale;
    deltat = fusion.deltatUpdate(); //this have to be done before calling the fusion update    //choose only one of these two:    //    fusion.MahonyUpdate(gx, gy, gz, ax, ay, az, deltat);  //mahony is suggested if there isn't the mag and the mcu is slow    fusion.MadgwickUpdate(gx, gy, gz, ax, ay, az, mx, my, mz, deltat);  //else use the magwick, it is slower but more accurate
    pitch = fusion.getPitchRadians();    roll = -fusion.getRollRadians();    //you could also use getRollRadians() etc    yaw = fusion.getYawRadians();
    Serial.print(pitch);    Serial.print("\t");    Serial.print(roll);    Serial.print("\t");    Serial.print(yaw);    Serial.print("\t");
    Serial.print(ax);    Serial.print("\t");    Serial.print(ay);    Serial.print("\t");    Serial.print(az);    Serial.print("\t");
    Serial.print(gx);    Serial.print("\t");    Serial.print(gy);    Serial.print("\t");    Serial.println(gz);  }}

将四元数的姿态信息用processing做可视化后，才发现现实的残酷。四元数能够很好地反映开发板的姿态信息，通过仰角也是能够获得开发板与水平面的夹角，可是这个仰角数据滞后严重啊！尝试旋转开发板，然后水平放置在桌面上，图中的立方体，是缓慢恢复到水平态的！而用手投球，发力时长是按毫秒计算的。这样的滞后，感觉没法算了。

更换思路：放弃通过加速度去计算速度的方法，换种思考角度，尝试通过机器学习来解决这个问题。传感器获得的信息是条件，计算出篮球脱手的速度和角度是方法，是否命中篮筐是结果。既然方法这块弄起来有困难，那么就给足够多的条件和结果，用机器学习的方式去推算方法。

首先，收集几组投篮的数据，来明确条件的边界。上图中图像是将三个轴的加速度做了平方和的结果。可以看出平时的运动，三个轴加速度和基本为1，投篮时，加速度会变大，图像上有明显的突变。这里每个点之间是50ms（这里应该是串口影响，每次取样时间间隔近50ms，传感器的频率应该是110HZ，间隔应该在10ms才对），一次投篮的加速度差不多350ms左右。取加速度和为35作为阈值。开始训练，通过串口记录训练样本值。

这里保留影响投篮的相关参数：三轴加速度，三轴角速度。其它传感器信息都不要。以加速度和超过35的点作为中心点，前后一共取32个点的数据。当出现符合条件的情况，就通过串口把数据传送给电脑，记录下来作为训练数据。实际测量时发现有投完球后也会有数据出现，就再从时间上限定，出现符合条件的数据后1秒内不再接受数据。

因为需要接电脑收集数据，不方便使用篮球场收集数据（就是因为宅，且不会打篮球，而且也没有篮球）。就改用孩子的米奇玩偶，在室内的墙上做一个标记位，每次击中标记位即为投中。将开发板绑在手背上，用usb延长线与电脑连接。米奇每击中标记位，就记录一次数据，每组数据均是32*6=192位的数据。应该记录100次以上的，但是为了测试（宅男属性，体力不支）收集到了62组数据就停止了。

2.38      -3.86      3.05      109.99      -37.11      29.54      2.42      -4      3.15      119.2      -33.75      38.09      2.26      -4      2.83      119.51      -17.82      69.82      2.22      -3.68      2.02      107.6      -6.71      116.46      2.27      -3.05      1.59      89.66      8.48      171.02      1.95      -2.83      1.54      58.29      30.58      244.2      1.38      -2.61      1.4      2.14      63.96      316.47      1.44      -1.88      1.55      -91.19      105.35      365.23      1.13      -2.08      1.97      -211.43      149.72      428.16      0.98      -2.13      2.57      -268.55      260.19      506.71      -0.88      -2.48      2.99      -222.29      360.23      612.37      -3.98      -1.05      1.76      -105.83      356.38      700.93      -4      1.42      0.04      -56.15      430.66      745.18      -4      3.43      -0.39      -177.98      635.07      672.24      -4      4      -1.42      -141.91      619.14      616.46      -4      4      -3.21      -288.51      595.58      332.7      -4      4      -4      -406.19      628.66      125.49      -4      4      -4      -522.77      520.69      56.21      -4      0.99      -4      -439.39      152.71      -34.48      -3.48      -0.33      -3.64      -139.1      -85.88      -300.78      -2.19      -1.99      -3.22      127.01      -62.5      -454.53      -1.21      -3.13      -2.13      189.27      -53.65      -430.85      -0.34      -3.99      -0.93      112.3      -162.72      -322.51      0.73      -4      -0.11      41.26      -242.37      -220.83      0.38      -3.58      -0.02      43.4      -222.53      -187.74      0.73      -3.5      0.26      86      -164.92      -133.06      0.79      -2.52      -0.04      78.37      -105.16      -90.64      1.02      -2.03      0.01      19.78      -78.06      -57.62      0.88      -1.59      0.17      -48.77      -58.78      -32.35      0.68      -1.51      0.28      -69.15      -47.24      -1.16      0.37      -1.23      0.37      -62.87      -32.04      5.86      0.55      -0.65      0.19      -43.46      -17.09      0.851.79      -3.44      3.57      128.72      -54.99      59.2      1.82      -3.52      3.43      130.62      -53.96      81.91      1.94      -3.5      3.45      115.05      -29.79      125.55      1.9      -3.65      3.4      61.77      12.21      194.64      1.87      -4      3.33      -20.75      73.36      280.46      1.53      -3.98      3.38      -89.05      135.44      368.29      1.42      -3.27      3.48      -136.35      213.2      437.19      0.76      -2.55      3.62      -207.34      332.76      481.26      -0.45      -2.58      3.97      -256.71      486.88      517.33      -2.33      -2.91      3.57      -205.2      605.41      559.94      -3.98      -2.39      2.84      -46.57      654.91      608.52      -4      0.12      1.42      178.59      715.09      603.09      -4      2.68      1.49      338.62      791.2      533.69      -4      4      0.2      280.09      792.97      479.37      -4      4      0.35      21.12      810.55      411.07      -4      4      -2.11      -226.87      810.73      325.99      -4      4      -2.85      -437.19      638.37      226.56      -4      3.81      -3.64      -456.73      539.25      193.05      -4      1.9      -4      -234.99      430.18      76.11      -3.3      1.32      -3.99      54.44      220.95      -187.19      -2.56      1.23      -3.42      192.75      78.49      -420.78      -2.06      -0.63      -3      201.29      -2.93      -438.6      -1.68      -1.91      -2.19      77.45      -41.56      -441.47      -0.6      -1.38      -1.06      -57.92      -40.95      -428.16      -0.6      -3.58      -0.57      -74.04      -61.77      -305.85      0.2      -2.99      -0.15      -86.43      -70.31      -247.68      0.61      -2.22      -0.05      -79.71      -80.69      -155.76      0.66      -2.09      0.04      -70.31      -90.27      -47.12      0.71      -1.51      -0.1      -83.68      -91.13      -2.38      0.8      -1.3      -0.23      -98.75      -86.55      20.51      0.74      -1.21      -0.36      -106.75      -86.67      50.66      0.76      -0.95      -0.25      -119.08      -52.61      60.972.62      -2.9      2.84      88.93      -42.24      47.3      2.69      -3.39      3.11      126.16      -76.9      21.73      3.18      -3.64      3.33      138.73      -68.54      -17.03      3.22      -4      3.3      120.42      -40.16      -19.29      3.09      -4      2.79      92.1      -30.03      31.13      3.16      -4      2.08      70.19      0.79      73.97      2.97      -3.94      2.03      57.13      47.49      137.33      2.68      -3.6      1.6      51.57      85.14      225.22      2      -3.15      1.17      -6.29      139.1      323.67      1.16      -2.27      0.9      -102.48      206.42      446.66      -0.19      -1.81      0.85      -209.66      259.64      559.69      -3.03      0.16      0.19      -266.17      277.34      693.79      -4      1.68      0.02      -266.54      327.45      801.7      -4      3.24      -0.46      -89.54      438.35      758.73      -4      4      -0.79      94.42      535.64      518.8      -4      4      -1.78      134.22      605.1      240.84      -4      4      -2.27      13.67      581.97      -32.35      -4      4      -3.57      -206.18      424.5      -214.17      -4      2.76      -2.58      -392.88      258.06      -284.48      -3.9      -1.8      -2.07      -330.2      86.43      -265.81      -2.04      -3.35      -2.45      -90.52      -12.15      -273.56      -1.13      -2.71      -2.49      96.19      24.6      -300.72      -0.69      -2.12      -2.23      146      32.78      -275.51      -0.13      -1.81      -1.6      102.97      -58.96      -230.96      0.07      -1.88      -0.93      51.15      -168.15      -180.24      0.38      -2.19      -0.32      26.25      -198.18      -147.22      0.57      -2.25      0.07      2.14      -167.18      -141.48      1.04      -2.07      0.34      -25.94      -129.09      -122.62      1.33      -1.93      0.56      -45.41      -96.5      -54.81      0.65      -2.34      0.93      -58.96      -63.05      31.98      0.98      -1.52      0.9      -80.99      -45.78      56.4      0.83      -1.08      0.66      -90.64      -24.11      71.17

再在屋顶标记一个目标点，做为用力过猛的点，继续收集数据，收集到了60组数据。

0.09      0.99      -0.73      -2.26      37.78      51.51      0.08      0.99      -0.78      0.85      37.41      42.3      0.07      0.97      -0.81      1.1      33.69      35.58      0      1.05      -0.85      0.12      28.87      31.49      -0.1      1.07      -0.88      -0.49      24.54      29.54      -0.2      1.12      -0.88      0.67      23.44      31.13      -0.27      1.15      -0.9      1.34      28.14      36.01      -0.3      1.18      -0.95      -3.66      34.06      44.19      -0.34      1.34      -1.02      -13.49      38.76      50.17      -0.4      1.54      -1.13      -18.98      43.95      51.45      -0.54      1.76      -1.22      -10.68      50.17      47.55      -0.78      1.97      -1.36      4.27      59.14      35.83      -1.16      2.49      -1.54      11.17      67.26      22.46      -1.88      2.92      -1.63      14.89      69.09      10.19      -2.71      3.44      -2.27      8.24      85.33      7.39      -2.96      4      -3.38      -43.33      114.75      8.06      -2.57      4      -4      -68.3      140.93      -1.59      -1.87      4      -4      -39.18      125.18      -47.61      -1.21      3.75      -4      -9.7      71.72      -102.11      -0.84      3.41      -4      -19.78      -32.1      -191.41      -1.33      3.02      -3.61      -10.44      -113.77      -261.78      -0.43      2.38      -3.81      -47.73      -158.75      -331.85      0.03      2.24      -3.86      -142.21      -238.83      -362.61      0.19      1.6      -3.75      -218.08      -355.65      -402.53      0.25      1.03      -3.3      -304.38      -483.89      -446.17      0.36      -0.14      -2.38      -379.7      -635.99      -511.47      -1.18      -0.94      -0.63      -407.04      -847.29      -627.81      -4      -0.38      1.99      -304.02      -1102.17      -790.1      -4      -1.76      4      -14.89      -1289.79      -969.79      -4      -4      4      558.23      -1196.35      -800.78      -4      -4      3.84      759.03      -510.93      -295.84      -3.82      -4      2.76      575.32      -178.83      13.18-0.72      0.84      -1.08      -0.18      18.19      27.22      -0.69      0.9      -1.11      -1.89      22.64      19.53      -0.7      0.93      -1.11      0.37      27.22      8.36      -0.75      0.94      -1.15      5.55      29.24      -3.17      -0.79      0.91      -1.19      9.22      32.23      -13.31      -0.87      0.88      -1.22      12.08      35.89      -21.42      -0.91      0.88      -1.29      12.94      39.98      -27.95      -1      0.91      -1.37      9.46      44.43      -33.14      -1.11      1.01      -1.47      4.39      48.4      -37.29      -1.33      1.14      -1.57      0.61      45.47      -42.36      -1.71      1.37      -1.74      -0.98      36.93      -48.28      -2.15      1.59      -2.02      1.1      30.58      -53.71      -2.54      1.83      -2.48      0.12      39.43      -61.1      -3.02      2.22      -3.16      -0.92      60.42      -81.79      -3.54      2.64      -3.94      5.92      85.02      -117      -3.67      2.63      -4      16.48      111.27      -156.01      -3.39      2.76      -4      11.84      125.18      -198.43      -2.85      2.68      -4      17.21      108.28      -254.03      -2.21      2.35      -4      26.31      46.57      -322.88      -1.67      2.28      -4      13.98      -37.17      -381.04      -1.61      2.65      -3.99      -20.45      -147.34      -417.48      -1.35      2.3      -3.78      -51.03      -281.56      -453.86      -1.38      3.13      -3.15      -74.46      -424.8      -493.29      -0.88      1.09      -2.76      -58.53      -599.06      -550.11      -1.33      -0.3      -1.79      -76.72      -798.03      -646.42      -2.64      -0.94      -0.48      -171.94      -1031.07      -834.41      -4      -1.61      0.28      -188.78      -1302.12      -975.34      -4      -3.12      0.64      -229.37      -1593.69      -1125.61      -4      -4      4      -267.7      -1458.44      -1121.4      -4      -4      4      -254.15      -523.25      -570.13      -4      -4      4      128.48      -244.81      -61.16      -4      -4      4      302.8      -152.77      280.52-0.75      1.02      -0.99      -14.83      15.14      41.38      -0.85      1      -0.99      -11.05      13.31      33.63      -0.85      0.99      -0.99      -10.25      18.74      23.19      -0.81      1.01      -1      -9.52      23.25      12.45      -0.86      1.01      -1      -7.81      25.15      3.54      -0.92      1.04      -1.01      -6.9      27.89      -3.36      -0.99      1.06      -1.03      -6.84      31.25      -9.58      -1.05      1.07      -1.06      -7.87      36.32      -15.63      -1.07      1.1      -1.12      -11.54      43.09      -18.92      -1.1      1.16      -1.17      -17.64      50.05      -21.61      -1.25      1.28      -1.24      -19.1      55.79      -26.73      -1.55      1.47      -1.36      -17.33      58.35      -35.52      -2.09      1.91      -1.63      -14.4      59.69      -47.55      -2.89      2.5      -2      -10.01      60.85      -57.5      -3.58      2.87      -2.73      -10.86      64.33      -65.25      -3.8      2.92      -3.47      -22.64      72.2      -84.96      -3.44      2.8      -3.99      -34.97      83.01      -130.25      -2.96      2.74      -4      -35.58      77.76      -202.15      -2.7      2.19      -4      -29.72      34.85      -296.2      -2.42      2.44      -4      -40.65      -40.65      -396.67      -2.21      1.92      -3.96      -38.15      -138.12      -465.76      -2.3      2.57      -3.42      -43.88      -248.9      -514.59      -1.83      2.2      -2.91      -85.33      -356.81      -558.17      -1.44      1.93      -2      -168.4      -484.5      -619.63      -1.27      1.28      -1.42      -253.05      -636.84      -660.16      -1.2      2.03      -0.7      -346.37      -820.8      -734.37      -3.96      -0.65      -0.56      -350.52      -1091.98      -888.18      -4      -0.75      0.63      -308.84      -1445.92      -942.63      -4      0.85      2.89      -228.33      -1602.97      -1137.27      -4      -4      4      60.24      -1350.28      -1002.2      -4      -4      4      185.12      -460.94      -569.82      -4      -4      4      348.94      -176.76      -164.79

科学的办法应该是将传感器固定在手上，然后真实地去场地投篮，记录成功与失败的数据，要想细分，失败的数据还要进一步对失败原因进行标记，将每种情况进行分类。这里就先在屋中用测试数据代替了。仅仅分为投中和用力过猛两类。

收集到这些数据条件后，对数据对应的结果进行标记。投中的标记为0，用力过猛的标记为1.数据标记后，做归一处理，通过查询，可以得知，加速度的量程为【-4，4】，角速度为【-2000，2000】，所以引入这样一个矩阵 dealzero=(4,4,4,2000,2000,2000)*32，所有的数据都与它除，就能将数据点，归到【-1，1】之间了。这里还发现一个问题，少量的加速度到了4这个值了，也就是说投球过程中，已经超出量程了，感觉【-4，4】这个量程不是太合适了。

#!/usr/bin/env python# -*- coding: utf-8 -*-# @Time    : 2021/04/21 14:25# @Author  : aramy# @File    : nano33    deal_trandate.py#处理训练数据import numpy as npimport pandas as pdimport tensorflow as tfimport tensorflow.keras as kerasdealzero=(4,4,4,2000,2000,2000)*32def processDate(file,y_val):    openD = pd.read_csv(file, header=None)    dataX = np.array(openD)    if dataX.shape[1]>192:        dataX=dataX[:,:-1]    dataX=dataX/dealzero    dataY=np.ones((dataX.shape[0],))*y_val    return dataX,dataY
def loaddata():    dataX0,dataY0 = processDate('../res/succ.csv',0)    dataX1,dataY1 = processDate('../res/fail.csv', 1)
    dataX = np.concatenate((dataX0, dataX1), axis=0)    dataY=np.append(dataY0,dataY1)    # print(dataX,dataY)    permutationTrain = np.random.permutation(dataX.shape[0])    # # print(permutationTrain)    dataX = dataX[permutationTrain]    dataY = dataY[permutationTrain]    vfoldSize = int(dataX.shape[0] / 100 * 20)    xTest = dataX[0:vfoldSize]    yTest = dataY[0:vfoldSize]    xTrain = dataX[vfoldSize:dataX.shape[0]]    yTrain = dataY[vfoldSize:dataY.shape[0]]
    return xTest,yTest,xTrain,yTrain

if __name__ == '__main__':    xTest,yTest,xTrain,yTrain=loaddata()    print( xTest,yTest,xTrain,yTrain)
    model = keras.Sequential()    model.add(keras.layers.Dense(32, input_shape=(6*32,), activation='relu'))    model.add(keras.layers.Dense(16, activation='relu'))    model.add(keras.layers.Dense(2, activation='softmax'))
    adam = keras.optimizers.Adam(0.000005)    model.compile(loss='sparse_categorical_crossentropy', optimizer=adam, metrics=['sparse_categorical_accuracy'])    model.summary()
    history = model.fit(xTrain, yTrain, batch_size=1, validation_data=(xTest, yTest), epochs=3000, verbose=1)    converter = tf.lite.TFLiteConverter.from_keras_model(model)    tflite_model = converter.convert()
    open("model", "wb").write(tflite_model)

训练模型：将归一后的数据用随机数的方式打乱，保证成功和用力过猛的数据均匀混合，取80%做为训练数据，20%做为验证数据。

我们使用的是Sequential模型，Sequential模型形象地解释就是一个漏斗，数据只能从上层往下层流动，我们可以在这个漏斗中堆叠各种计算的层，以丰富这个模型的结构。这里使用了三层的神经网络。使用softmax激活函数。这部分自己还不是太懂，只能是照猫画虎。使用0.000005做为损失函数，训练3000次。在我的电脑上，大概是5分钟，训练完成，得到一个model的文件。使用命令：xxd -i model >> model.h 就生成了Arduino下能用的model.h文件了。

使用训练结果：训练完成后，得到的model.h文件就要需要的模型了。我的理解是在Arudino收集到的运动数据来和模型作比较，看动作数据与之前训练中的特征值相似程度，用来判断当前动作的属性。在Arduino中增加tensor处理的框架，保证每次三轴加速度和超过35就触发tensor的判断，判断当前动作投篮是否成功。若投篮成功概率超过0.8，则在oled屏幕上显示“success”。若投篮用力过大的概率超过0.8，则显示“Too Hard”。将投篮次数、各自的概率显示在oled屏幕上。

#include#include "Buff.h"#include#include#include#include#include#include#include "model.h"#includeAdafruit_SSD1306 display(128, 64, &Wire, NULL);

float ax, ay, az, gx, gy, gz;float buff[actionnumber * 6];uint32_t tim[actionnumber];uint32_t rectime = 0;const float zero[3] = {0.12620776, -0.360400501, -0.518010013};   //归零Buff actionBuf;
tflite::MicroErrorReporter tflErrorReporter;tflite::AllOpsResolver tflOpsResolver;const tflite::Model* tflModel = nullptr;tflite::MicroInterpreter* tflInterpreter = nullptr;TfLiteTensor* tflInputTensor = nullptr;TfLiteTensor* tflOutputTensor = nullptr;constexpr int tensorArenaSize = 8 * 1024;byte tensorArena[tensorArenaSize];const char* ACTION[] = {  "SUCCESS",  "TOO HARD"};#define ACT_NUM (sizeof(ACTION) / sizeof(ACTION[0]))float answer[sizeof(ACTION) / sizeof(ACTION[0])];int action_num=0;         //投篮次数
void setup() {  Serial.begin(115200);  tflModel = tflite::GetModel(model);  if (tflModel->version() != TFLITE_SCHEMA_VERSION) {    Serial.println("Model schema mismatch!");    while (1);  }  tflInterpreter = new tflite::MicroInterpreter(tflModel, tflOpsResolver, tensorArena, tensorArenaSize, &tflErrorReporter);  tflInterpreter->AllocateTensors();  tflInputTensor = tflInterpreter->input(0);  tflOutputTensor = tflInterpreter->output(0);  delay(1000);  display.begin(SSD1306_SWITCHCAPVCC, 0x3C);    //初始化oled  display.clearDisplay();  display.display();  if (!IMU.begin()) {    Serial.println("Failed to initialize IMU!");    while (1);  }  delay(3000);
}
void loop() {  if (IMU.accelerationAvailable() && IMU.gyroscopeAvailable() ) {    IMU.readAcceleration(ax, ay, az);    IMU.readGyroscope(gx, gy, gz);    actionBuf.insertnode(gx, gy, gz, ax, ay, az, millis());    if (actionBuf.isavailable()) {      actionBuf.getActionVal(buff, tim);      if (millis() - rectime > 1000) {        rectime = millis();        judgeAcion();     //判断是否投中        action_num++;        dispactioninfo();    //显示信息      }    }  }}//将收到的数据作推理，判断是否能进球void judgeAcion() {  float max = 0;  char pos;  //数据归一  for (int i = 0; i << span=""> actionnumber; i++) {    tflInputTensor->data.f[i * 6 + 0] = buff[i * 6 + 0] / 4;    tflInputTensor->data.f[i * 6 + 1] = buff[i * 6 + 1] / 4;    tflInputTensor->data.f[i * 6 + 2] = buff[i * 6 + 2] / 4;    tflInputTensor->data.f[i * 6 + 3] = buff[i * 6 + 3] / 2000;    tflInputTensor->data.f[i * 6 + 4] = buff[i * 6 + 4] / 2000;    tflInputTensor->data.f[i * 6 + 5] = buff[i * 6 + 5] / 2000;  }  TfLiteStatus invokeStatus = tflInterpreter->Invoke();  if (invokeStatus != kTfLiteOk) {    Serial.println("Invoke failed!");    while (1);    return;  }  for (int i = 0; i << span=""> ACT_NUM; i++) {    if (max << span=""> tflOutputTensor->data.f[i]) {      max = tflOutputTensor->data.f[i];      pos = i;    }    Serial.print(ACTION[i]);    Serial.print(": ");    Serial.print(tflOutputTensor->data.f[i], 6);    Serial.print("   ");    answer[i]=tflOutputTensor->data.f[i];  }  Serial.println();}
//显示结果信息void dispactioninfo(){  display.clearDisplay(); //清空屏幕  display.setCursor(0, 0); //设置起点  display.setTextSize(2); //设置字体  display.setTextColor(SSD1306_WHITE); //设置字体颜色  display.println(F("NO."));  display.setCursor(60, 0); //设置起点  display.println(action_num);    //显示次数  display.setTextSize(3); //设置字体  display.setCursor(0, 24); //设置起点  if(answer[0]>0.8){//投篮成功    display.println("SUCCESS");  }else if(answer[1]>0.8){    display.println("TOO HARD");  }else{    display.println("FAIL");  }  display.setTextSize(1); //设置字体  display.setCursor(0, 48); //设置起点  display.print(answer[0],3);  display.print(" , ");  display.print(answer[1],3);  display.display(); //刷新屏幕  delay(3000);}