PX4姿态解算磁偏补偿

PX4磁偏补偿

PX4 姿态解算中可以通过参数直接设置磁偏角，如果有GPS也可以通过GPS座标查询。

// px4 attitude_estimator_q.cpp - update_parameters
float mag_decl_deg = 0.0f;
param_get(_params_handles.mag_decl, &mag_decl_deg);
update_mag_declination(math::radians(mag_decl_deg));

if (_mag_decl_auto && gpos.eph < 20.0f && hrt_elapsed_time(&gpos.timestamp) < 1000000)
 {
    /* set magnetic declination automatically */
	update_mag_declination(math::radians(get_mag_declination(gpos.lat, gpos.lon)));
}

磁偏数据库

通过GPS座标查询磁偏角需要首先建立磁偏数据库，磁偏角查询
根据px4中设置的座标范围和划分精度，建立数据库，设置如下图。

/** set this always to the sampling in degrees for the table below */
static constexpr float SAMPLING_RES = 10.0f;
static constexpr float SAMPLING_MIN_LAT	= -60.0f;
static constexpr float SAMPLING_MAX_LAT	= 60.0f;
static constexpr float SAMPLING_MIN_LON	= -180.0f;
static constexpr float SAMPLING_MAX_LON	= 180.0f;

导出的数据库如下图，PX4中存储的是取整后的值，会发现两个表中有的数会有1度的差别，因为这个值是缓慢变化的。

磁偏计算

磁偏数据库建立完成之后就可以根据GPS座标查询到上图中对应的网格，PX4中采用双线性插值，即根据网格四个顶点的磁偏角计算座标所在点的磁偏角。

static float
get_table_data(float lat, float lon, const int8_t table[13][37])
{
	/*
	 * If the values exceed valid ranges, return zero as default
	 * as we have no way of knowing what the closest real value
	 * would be.
	 */
	if (lat < -90.0f || lat > 90.0f ||
	    lon < -180.0f || lon > 180.0f) {
		return 0.0f;
	}

	/* round down to nearest sampling resolution */
	float min_lat = floorf(lat / SAMPLING_RES) * SAMPLING_RES;
	float min_lon = floorf(lon / SAMPLING_RES) * SAMPLING_RES;

	/* find index of nearest low sampling point */
	unsigned min_lat_index = get_lookup_table_index(&min_lat, SAMPLING_MIN_LAT, SAMPLING_MAX_LAT);
	unsigned min_lon_index = get_lookup_table_index(&min_lon, SAMPLING_MIN_LON, SAMPLING_MAX_LON);

	const float data_sw = table[min_lat_index][min_lon_index];
	const float data_se = table[min_lat_index][min_lon_index + 1];
	const float data_ne = table[min_lat_index + 1][min_lon_index + 1];
	const float data_nw = table[min_lat_index + 1][min_lon_index];

	/* perform bilinear interpolation on the four grid corners */
	const float lat_scale = constrain((lat - min_lat) / SAMPLING_RES, 0.0f, 1.0f);
	const float lon_scale = constrain((lon - min_lon) / SAMPLING_RES, 0.0f, 1.0f);

	const float data_min = lon_scale * (data_se - data_sw) + data_sw;
	const float data_max = lon_scale * (data_ne - data_nw) + data_nw;

	return lat_scale * (data_max - data_min) + data_min;
}

对应算法如下图：

双线性插值，可理解为两次插值，分别插值出R1,R2, 再通过R1 R2插值出P.