accelerometer - 为 IMU 加速证明创建算法

Question

我尝试设计一种适用于自行车的 IMU 算法。我使用该算法来估算 IMU 的欧拉角。我只使用加速度计和陀螺仪来估计我的姿势。不幸的是，当我烘烤时，由于减速，我得到了错误的姿势估计。

这是我用来估计 IMU 的算法的一部分：

  float recipNorm;
  float halfvx, halfvy, halfvz;
  float halfex, halfey, halfez;
  float qa, qb, qc;

  // Compute feedback only if accelerometer measurement valid (avoids NaN in
  // accelerometer normalisation)
  if (!((ax == 0.0f) && (ay == 0.0f) && (az == 0.0f))) {

    // Normalise accelerometer measurement
    recipNorm = invSqrt(ax * ax + ay * ay + az * az);
    ax *= recipNorm;
    ay *= recipNorm;
    az *= recipNorm;

    // Estimated direction of gravity and vector perpendicular to magnetic flux
    halfvx = q1 * q3 - q0 * q2;
    halfvy = q0 * q1 + q2 * q3;
    halfvz = q0 * q0 - 0.5f + q3 * q3;

    // Error is sum of cross product between estimated and measured direction of
    // gravity
    halfex = (ay * halfvz - az * halfvy);
    halfey = (az * halfvx - ax * halfvz);
    halfez = (ax * halfvy - ay * halfvx);

    // Compute and apply integral feedback if enabled
    if (twoKi > 0.0f) {
      integralFBx +=
          twoKi * halfex * (1.0f / sampleFreq); // integral error scaled by Ki
      integralFBy += twoKi * halfey * (1.0f / sampleFreq);
      integralFBz += twoKi * halfez * (1.0f / sampleFreq);
      gx += integralFBx; // apply integral feedback
      gy += integralFBy;
      gz += integralFBz;
    } else {
      integralFBx = 0.0f; // prevent integral windup
      integralFBy = 0.0f;
      integralFBz = 0.0f;
    }

    // Apply proportional feedback
    gx += twoKp * halfex;
    gy += twoKp * halfey;
    gz += twoKp * halfez;
  }

  // Integrate rate of change of quaternion
  gx *= (0.5f * (1.0f / sampleFreq)); // pre-multiply common factors
  gy *= (0.5f * (1.0f / sampleFreq));
  gz *= (0.5f * (1.0f / sampleFreq));
  qa = q0;
  qb = q1;
  qc = q2;
  q0 += (-qb * gx - qc * gy - q3 * gz);
  q1 += (qa * gx + qc * gz - q3 * gy);
  q2 += (qa * gy - qb * gz + q3 * gx);
  q3 += (qa * gz + qb * gy - qc * gx);

  // Normalise quaternion
  recipNorm = invSqrt(q0 * q0 + q1 * q1 + q2 * q2 + q3 * q3);
  q0 *= recipNorm;
  q1 *= recipNorm;
  q2 *= recipNorm;
  q3 *= recipNorm;

使用该参数，我尝试修改：

#define sampleFreq (50f/2.f)      // sample frequency in Hz
#define twoKpDef (2.0f * 0.1f) // 2 * proportional gain
#define twoKiDef (2.0f * 1.f)  // 2 * integral gain

我无法从我的姿势估计中提取减速。每次我刹车时，都会出现错误的角度估计。

有人知道我的目的更有效的算法或设置参数的方式吗？

标签： accelerometergyroscopeimu