Odrive源码分析(六) 相关控制变量传递
本文记录下odrive源代码中相关控制模块之间变量的传递,这对理解odrive源代码至关重要。
通过前面文字的分析,odrive有两条数据链路,一条是通过中断进行实时的控制,另外一条是OS相关的操作,主要分析下中断内部的相关变量传递。
中断函数:TIM8_UP_TIM13_IRQHandler,,此中断又触发了软中断ControlLoop_IRQHandler。在ControlLoop_IRQHandler中,会进行数据的更新,主要是调用:odrv.control_loop_cb。
下面分析odrv.control_loop_cb函数内部进行的数据更新操作:
void ODrive::control_loop_cb(uint32_t timestamp) {
last_update_timestamp_ = timestamp;
n_evt_control_loop_++;
// TODO: use a configurable component list for most of the following things
MEASURE_TIME(task_times_.control_loop_misc) {
// Reset all output ports so that we are certain about the freshness of
// all values that we use.
// If we forget to reset a value here the worst that can happen is that
// this safety check doesn't work.
// TODO: maybe we should add a check to output ports that prevents
// double-setting the value.
for (auto& axis: axes) {
axis.acim_estimator_.slip_vel_.reset();
axis.acim_estimator_.stator_phase_vel_.reset();
axis.acim_estimator_.stator_phase_.reset();
axis.controller_.torque_output_.reset();
axis.encoder_.phase_.reset();
axis.encoder_.phase_vel_.reset();
axis.encoder_.pos_estimate_.reset();
axis.encoder_.vel_estimate_.reset();
axis.encoder_.pos_circular_.reset();
axis.motor_.Vdq_setpoint_.reset();
axis.motor_.Idq_setpoint_.reset();
axis.open_loop_controller_.Idq_setpoint_.reset();
axis.open_loop_controller_.Vdq_setpoint_.reset();
axis.open_loop_controller_.phase_.reset();
axis.open_loop_controller_.phase_vel_.reset();
axis.open_loop_controller_.total_distance_.reset();
axis.sensorless_estimator_.phase_.reset();
axis.sensorless_estimator_.phase_vel_.reset();
axis.sensorless_estimator_.vel_estimate_.reset();
}
uart_poll();
odrv.oscilloscope_.update();
}
for (auto& axis : axes) {
MEASURE_TIME(axis.task_times_.endstop_update) {
axis.min_endstop_.update();
axis.max_endstop_.update();
}
}
MEASURE_TIME(task_times_.control_loop_checks) {
for (auto& axis: axes) {
// look for errors at axis level and also all subcomponents
bool checks_ok = axis.do_checks(timestamp);
// make sure the watchdog is being fed.
bool watchdog_ok = axis.watchdog_check();
if (!checks_ok || !watchdog_ok) {
axis.motor_.disarm();
}
}
}
for (auto& axis: axes) {
// Sub-components should use set_error which will propegate to this error_
MEASURE_TIME(axis.task_times_.thermistor_update) {
axis.motor_.fet_thermistor_.update();
axis.motor_.motor_thermistor_.update();
}
MEASURE_TIME(axis.task_times_.encoder_update)
axis.encoder_.update();
}
// Controller of either axis might use the encoder estimate of the other
// axis so we process both encoders before we continue.
for (auto& axis: axes) {
MEASURE_TIME(axis.task_times_.sensorless_estimator_update)
axis.sensorless_estimator_.update();
MEASURE_TIME(axis.task_times_.controller_update) {
if (!axis.controller_.update()) { // uses position and velocity from encoder
axis.error_ |= Axis::ERROR_CONTROLLER_FAILED;
}
}
MEASURE_TIME(axis.task_times_.open_loop_controller_update)
axis.open_loop_controller_.update(timestamp);
MEASURE_TIME(axis.task_times_.motor_update)
axis.motor_.update(timestamp); // uses torque from controller and phase_vel from encoder
MEASURE_TIME(axis.task_times_.current_controller_update)
axis.motor_.current_control_.update(timestamp); // uses the output of controller_ or open_loop_contoller_ and encoder_ or sensorless_estimator_ or acim_estimator_
}
// Tell the axis threads that the control loop has finished
for (auto& axis: axes) {
if (axis.thread_id_) {
osSignalSet(axis.thread_id_, 0x0001);
}
}
get_gpio(odrv.config_.error_gpio_pin).write(odrv.any_error());
}
下面这一段reset,是让相关数据过期,防止下一周期使用到了旧的数据。
MEASURE_TIME(task_times_.control_loop_misc) {
// Reset all output ports so that we are certain about the freshness of
// all values that we use.
// If we forget to reset a value here the worst that can happen is that
// this safety check doesn't work.
// TODO: maybe we should add a check to output ports that prevents
// double-setting the value.
for (auto& axis: axes) {
axis.acim_estimator_.slip_vel_.reset();
axis.acim_estimator_.stator_phase_vel_.reset();
axis.acim_estimator_.stator_phase_.reset();
axis.controller_.torque_output_.reset();
axis.encoder_.phase_.reset();
axis.encoder_.phase_vel_.reset();
axis.encoder_.pos_estimate_.reset();
axis.encoder_.vel_estimate_.reset();
axis.encoder_.pos_circular_.reset();
axis.motor_.Vdq_setpoint_.reset();
axis.motor_.Idq_setpoint_.reset();
axis.open_loop_controller_.Idq_setpoint_.reset();
axis.open_loop_controller_.Vdq_setpoint_.reset();
axis.open_loop_controller_.phase_.reset();
axis.open_loop_controller_.phase_vel_.reset();
axis.open_loop_controller_.total_distance_.reset();
axis.sensorless_estimator_.phase_.reset();
axis.sensorless_estimator_.phase_vel_.reset();
axis.sensorless_estimator_.vel_estimate_.reset();
}
}
这一段是更新限位状态
for (auto& axis : axes) {
MEASURE_TIME(axis.task_times_.endstop_update) {
axis.min_endstop_.update();
axis.max_endstop_.update();
}
}
各种报错检查
MEASURE_TIME(task_times_.control_loop_checks) {
for (auto& axis: axes) {
// look for errors at axis level and also all subcomponents
bool checks_ok = axis.do_checks(timestamp);
// make sure the watchdog is being fed.
bool watchdog_ok = axis.watchdog_check();
if (!checks_ok || !watchdog_ok) {
axis.motor_.disarm();
}
}
}
mos管温度和编码器位置值更新 这里关注的重点是编码器值的更新,因为后续会用到更新的值,是后续顺利进行的基础。
for (auto& axis: axes) {
// Sub-components should use set_error which will propegate to this error_
MEASURE_TIME(axis.task_times_.thermistor_update) {
axis.motor_.fet_thermistor_.update();
axis.motor_.motor_thermistor_.update();
}
MEASURE_TIME(axis.task_times_.encoder_update)
axis.encoder_.update();
}
axis.encoder_.update()函数会更新encode里面以下几组重要的变量:
pos_estimate_counts_ : 单位:编码器累计位置值.
pos_estimate_ : 单位:圈数。
pos_cpr_counts_: 单位:编码器累计位置值,但是会控制在编码器分辨率范围内,比如0-4000。
vel_estimate_counts_: 单位:编码器变化速度。
phase_: 电角度。
phase_vel_: 电角度速度。
接下来调用
// Controller of either axis might use the encoder estimate of the other
// axis so we process both encoders before we continue.
for (auto& axis: axes) {
MEASURE_TIME(axis.task_times_.sensorless_estimator_update)
axis.sensorless_estimator_.update();
MEASURE_TIME(axis.task_times_.controller_update) {
if (!axis.controller_.update()) { // uses position and velocity from encoder
axis.error_ |= Axis::ERROR_CONTROLLER_FAILED;
}
}
MEASURE_TIME(axis.task_times_.open_loop_controller_update)
axis.open_loop_controller_.update(timestamp);
MEASURE_TIME(axis.task_times_.motor_update)
axis.motor_.update(timestamp); // uses torque from controller and phase_vel from encoder
MEASURE_TIME(axis.task_times_.current_controller_update)
axis.motor_.current_control_.update(timestamp); // uses the output of controller_ or open_loop_contoller_ and encoder_ or sensorless_estimator_ or acim_estimator_
}
axis.controller_.update()会使用到encoder值对控制变量进行更新,在闭环开始时,已经对变量进行了链接:
controller_.pos_estimate_linear_src_.connect_to(&ax->encoder_.pos_estimate_);
controller_.vel_estimate_src_.connect_to(&ax->encoder_.vel_estimate_);
controller更新时会使用到外界给定的输入,比如位置控制会使用到input_pos_,速度控制会使用到input_vel__等。
然后计算后会对以下关键变量进行更新:
torque_output_ : 力矩输出值
然后来到了最关键的更新函数: axis.motor_.update(timestamp);
此函数会用到controller更新时计算的变量torque_output_:
motor_.torque_setpoint_src_.connect_to(&controller_.torque_output_);
Motor::update会计算出以下重要的变量:
Idq_setpoint_ : 电流环控制DQ
Vdq_setpoint_ : 电压控制DQ.
接下来就是FOC的更新了,调用 axis.motor_.current_control_.update(timestamp),实际上是调用的FieldOrientedController::update。
这里主要就是把上面计算过程中的变量传进FOC对象中:
void FieldOrientedController::update(uint32_t timestamp) {
CRITICAL_SECTION() {
ctrl_timestamp_ = timestamp;
enable_current_control_ = enable_current_control_src_;
Idq_setpoint_ = Idq_setpoint_src_.present();
Vdq_setpoint_ = Vdq_setpoint_src_.present();
phase_ = phase_src_.present();
phase_vel_ = phase_vel_src_.present();
}
}
到了这里,基本的计算就结束了(odrv.control_loop_cb函数).
上面的计算完成后,接下来就是相电流标定和SVPWM更新了。
motors[0].dc_calib_cb(timestamp + TIM_1_8_PERIOD_CLOCKS * (TIM_1_8_RCR + 1) - TIM1_INIT_COUNT, current0);
motors[1].dc_calib_cb(timestamp + TIM_1_8_PERIOD_CLOCKS * (TIM_1_8_RCR + 1), current1);
motors[0].pwm_update_cb(timestamp + 3 * TIM_1_8_PERIOD_CLOCKS * (TIM_1_8_RCR + 1) - TIM1_INIT_COUNT);
motors[1].pwm_update_cb(timestamp + 3 * TIM_1_8_PERIOD_CLOCKS * (TIM_1_8_RCR + 1));
综上分析:基本上外界(上位机)传送目标位置或速度或力矩到controller模块,controller模块计算成变量torque,然后由motor模块计到DQ空间,之后就可以用FOC的控制流程进行反park变换,最终通过SVPWM控制三相马鞍波形输出实现电机运转,电机运转时又通过ADC采集电流进行Clark变换和Park变换得到测量的DQ,通过controller模块的PID运算实现整个系统的闭环,整个系统还是非常严谨的。