mirror of https://github.com/ArduPilot/ardupilot
424 lines
16 KiB
C++
424 lines
16 KiB
C++
/*
|
|
* control.cpp
|
|
* Copyright (C) Leonard Hall 2020
|
|
*
|
|
* This file is free software: you can redistribute it and/or modify it
|
|
* under the terms of the GNU General Public License as published by the
|
|
* Free Software Foundation, either version 3 of the License, or
|
|
* (at your option) any later version.
|
|
*
|
|
* This file is distributed in the hope that it will be useful, but
|
|
* WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
|
|
* See the GNU General Public License for more details.
|
|
*
|
|
* You should have received a copy of the GNU General Public License along
|
|
* with this program. If not, see <http://www.gnu.org/licenses/>.
|
|
*/
|
|
|
|
/*
|
|
* this module provides common controller functions
|
|
*/
|
|
#include "AP_Math.h"
|
|
#include "vector2.h"
|
|
#include "vector3.h"
|
|
#include <AP_InternalError/AP_InternalError.h>
|
|
|
|
// control default definitions
|
|
#define CONTROL_TIME_CONSTANT_RATIO 4.0f // time constant to ensure stable kinimatic path generation
|
|
|
|
// update_vel_accel - single axis projection of velocity, vel, forwards in time based on a time step of dt and acceleration of accel.
|
|
// the velocity is not moved in the direction of limit if limit is not set to zero
|
|
void update_vel_accel(float& vel, float accel, float dt, float limit)
|
|
{
|
|
const float delta_vel = accel * dt;
|
|
if (!is_positive(delta_vel * limit)){
|
|
vel += delta_vel;
|
|
}
|
|
}
|
|
|
|
// update_pos_vel_accel - single axis projection of position and velocity, pos and vel, forwards in time based on a time step of dt and acceleration of accel.
|
|
// the position and velocity is not moved in the direction of limit if limit is not set to zero
|
|
void update_pos_vel_accel(float& pos, float& vel, float accel, float dt, float limit)
|
|
{
|
|
// move position and velocity forward by dt if it does not increase error when limited.
|
|
float delta_pos = vel * dt + accel * 0.5f * sq(dt);
|
|
if (!is_positive(delta_pos * limit)){
|
|
pos += delta_pos;
|
|
}
|
|
|
|
update_vel_accel(vel, accel, dt, limit);
|
|
}
|
|
|
|
// update_vel_accel - dual axis projection of position and velocity, pos and vel, forwards in time based on a time step of dt and acceleration of accel.
|
|
// the velocity is not moved in the direction of limit if limit is not set to zero
|
|
void update_vel_accel_xy(Vector2f& vel, const Vector2f& accel, float dt, Vector2f limit)
|
|
{
|
|
// increase velocity by acceleration * dt if it does not increase error when limited.
|
|
Vector2f delta_vel = accel * dt;
|
|
if (!is_zero(limit.length_squared())) {
|
|
// zero delta_vel if it will increase the velocity error
|
|
limit.normalize();
|
|
if (is_positive(delta_vel * limit)) {
|
|
delta_vel.zero();
|
|
}
|
|
}
|
|
vel += delta_vel;
|
|
}
|
|
|
|
// update_pos_vel_accel - dual axis projection of position and velocity, pos and vel, forwards in time based on a time step of dt and acceleration of accel.
|
|
// the position and velocity is not moved in the direction of limit if limit is not set to zero
|
|
void update_pos_vel_accel_xy(Vector2f& pos, Vector2f& vel, const Vector2f& accel, float dt, Vector2f limit)
|
|
{
|
|
// move position and velocity forward by dt.
|
|
Vector2f delta_pos = vel * dt + accel * 0.5f * sq(dt);
|
|
|
|
if (!is_zero(limit.length_squared())) {
|
|
// zero delta_vel if it will increase the velocity error
|
|
limit.normalize();
|
|
if (is_positive(delta_pos * limit)) {
|
|
delta_pos.zero();
|
|
}
|
|
}
|
|
|
|
pos += delta_pos;
|
|
|
|
update_vel_accel_xy(vel, accel, dt, limit);
|
|
}
|
|
|
|
/* shape_accel calculates a jerk limited path from the current acceleration to an input acceleration.
|
|
The function takes the current acceleration and calculates the required jerk limited adjustment to the acceleration for the next time dt.
|
|
The kinematic path is constrained by :
|
|
acceleration limits - accel_min, accel_max,
|
|
time constant - tc.
|
|
The time constant defines the acceleration error decay in the kinematic path as the system approaches constant acceleration.
|
|
The time constant also defines the time taken to achieve the maximum acceleration.
|
|
The time constant must be positive.
|
|
The function alters the variable accel to follow a jerk limited kinematic path to accel_input
|
|
*/
|
|
void shape_accel(const float accel_input, float& accel,
|
|
const float accel_min, const float accel_max,
|
|
const float tc, const float dt)
|
|
{
|
|
// sanity check tc
|
|
if (!is_positive(tc)) {
|
|
return;
|
|
}
|
|
|
|
// Calculate time constants and limits to ensure stable operation
|
|
const float KPa = 1.0 / tc;
|
|
|
|
float jerk_max = 0.0;
|
|
if (is_negative(accel_min) && is_positive(accel_max)){
|
|
jerk_max = MAX(-accel_min, accel_max) * KPa;
|
|
}
|
|
|
|
// jerk limit acceleration change
|
|
float accel_delta = accel_input - accel;
|
|
if (is_positive(jerk_max)) {
|
|
accel_delta = constrain_float(accel_delta, -jerk_max * dt, jerk_max * dt);
|
|
}
|
|
accel += accel_delta;
|
|
|
|
// limit acceleration to accel_max
|
|
if (is_negative(accel_min) && is_positive(accel_max)){
|
|
accel = constrain_float(accel, accel_min, accel_max);
|
|
}
|
|
}
|
|
|
|
// 2D version
|
|
void shape_accel_xy(const Vector2f& accel_input, Vector2f& accel,
|
|
float accel_max, float tc, float dt)
|
|
{
|
|
// sanity check tc
|
|
if (!is_positive(tc)) {
|
|
return;
|
|
}
|
|
|
|
// Calculate time constants and limits to ensure stable operation
|
|
const float KPa = 1.0 / tc;
|
|
const float jerk_max = accel_max * KPa;
|
|
|
|
// jerk limit acceleration change
|
|
Vector2f accel_delta = accel_input - accel;
|
|
if (is_positive(jerk_max)) {
|
|
accel_delta.limit_length(jerk_max * dt);
|
|
}
|
|
accel = accel + accel_delta;
|
|
|
|
// limit acceleration to accel_max
|
|
if (is_negative(accel_max)) {
|
|
// we may want to allow this for some applications but call error for now.
|
|
INTERNAL_ERROR(AP_InternalError::error_t::invalid_arg_or_result);
|
|
} else if (is_positive(accel_max)) {
|
|
accel.limit_length(accel_max);
|
|
}
|
|
}
|
|
|
|
|
|
/* shape_vel_accel and shape_vel_xy calculate a jerk limited path from the current position, velocity and acceleration to an input velocity.
|
|
The function takes the current position, velocity, and acceleration and calculates the required jerk limited adjustment to the acceleration for the next time dt.
|
|
The kinematic path is constrained by :
|
|
maximum velocity - vel_max,
|
|
maximum acceleration - accel_max,
|
|
time constant - tc.
|
|
The time constant defines the acceleration error decay in the kinematic path as the system approaches constant acceleration.
|
|
The time constant also defines the time taken to achieve the maximum acceleration.
|
|
The time constant must be positive.
|
|
The function alters the variable accel to follow a jerk limited kinematic path to vel_input and accel_input
|
|
The accel_max limit can be removed by setting it to zero.
|
|
*/
|
|
void shape_vel_accel(const float vel_input1, const float accel_input,
|
|
const float vel, float& accel,
|
|
const float vel_min, const float vel_max,
|
|
const float accel_min, const float accel_max,
|
|
const float tc, const float dt)
|
|
{
|
|
// sanity check tc
|
|
if (!is_positive(tc)) {
|
|
return;
|
|
}
|
|
|
|
// Calculate time constants and limits to ensure stable operation
|
|
const float KPa = 1.0 / tc;
|
|
|
|
// we are changing vel_input, but don't want the change in the caller
|
|
float vel_input = vel_input1;
|
|
|
|
// limit velocity to vel_max
|
|
if (is_negative(vel_min) && is_positive(vel_max)){
|
|
vel_input = constrain_float(vel_input, vel_min, vel_max);
|
|
}
|
|
|
|
// velocity error to be corrected
|
|
float vel_error = vel_input - vel;
|
|
|
|
// acceleration to correct velocity
|
|
float accel_target = vel_error * KPa;
|
|
|
|
// velocity correction with input velocity
|
|
accel_target += accel_input;
|
|
|
|
shape_accel(accel_target, accel, accel_min, accel_max, tc, dt);
|
|
}
|
|
|
|
// 2D version
|
|
void shape_vel_accel_xy(const Vector2f &vel_input1, const Vector2f& accel_input,
|
|
const Vector2f& vel, Vector2f& accel,
|
|
const float vel_max, const float accel_max, const float tc, const float dt)
|
|
{
|
|
// sanity check tc
|
|
if (!is_positive(tc)) {
|
|
return;
|
|
}
|
|
Vector2f vel_input = vel_input1;
|
|
|
|
// limit velocity to vel_max
|
|
if (is_negative(vel_max)) {
|
|
INTERNAL_ERROR(AP_InternalError::error_t::invalid_arg_or_result);
|
|
} else if (is_positive(vel_max)) {
|
|
vel_input.limit_length(vel_max);
|
|
}
|
|
|
|
// Calculate time constants and limits to ensure stable operation
|
|
const float KPa = 1.0 / tc;
|
|
|
|
// velocity error to be corrected
|
|
const Vector2f vel_error = vel_input - vel;
|
|
|
|
// acceleration to correct velocity
|
|
Vector2f accel_target = vel_error * KPa;
|
|
accel_target += accel_input;
|
|
|
|
shape_accel_xy(accel_target, accel, accel_max, tc, dt);
|
|
}
|
|
|
|
/* shape_pos_vel_accel calculate a jerk limited path from the current position, velocity and acceleration to an input position and velocity.
|
|
The function takes the current position, velocity, and acceleration and calculates the required jerk limited adjustment to the acceleration for the next time dt.
|
|
The kinematic path is constrained by :
|
|
maximum velocity - vel_max,
|
|
maximum acceleration - accel_max,
|
|
time constant - tc.
|
|
The time constant defines the acceleration error decay in the kinematic path as the system approaches constant acceleration.
|
|
The time constant also defines the time taken to achieve the maximum acceleration.
|
|
The time constant must be positive.
|
|
The function alters the variable accel to follow a jerk limited kinematic path to pos_input, vel_input and accel_input
|
|
The vel_max, vel_correction_max, and accel_max limits can be removed by setting the desired limit to zero.
|
|
*/
|
|
void shape_pos_vel_accel(const float pos_input, const float vel_input, const float accel_input,
|
|
const float pos, const float vel, float& accel,
|
|
const float vel_correction_max, const float vel_min, const float vel_max,
|
|
const float accel_min, const float accel_max, const float tc, const float dt)
|
|
{
|
|
// sanity check tc
|
|
if (!is_positive(tc)) {
|
|
return;
|
|
}
|
|
|
|
// Calculate time constants and limits to ensure stable operation
|
|
const float KPv = 1.0 / (CONTROL_TIME_CONSTANT_RATIO*tc);
|
|
const float accel_tc_max = accel_max*(1-1.0f/CONTROL_TIME_CONSTANT_RATIO);
|
|
|
|
// position error to be corrected
|
|
float pos_error = pos_input - pos;
|
|
|
|
// velocity to correct position
|
|
float vel_target = sqrt_controller(pos_error, KPv, accel_tc_max, dt);
|
|
|
|
// limit velocity correction to vel_correction_max
|
|
if (is_positive(vel_correction_max)) {
|
|
vel_target = constrain_float(vel_target, -vel_correction_max, vel_correction_max);
|
|
}
|
|
|
|
// velocity correction with input velocity
|
|
vel_target += vel_input;
|
|
|
|
shape_vel_accel(vel_target, accel_input, vel, accel, vel_min, vel_max, accel_min, accel_max, tc, dt);
|
|
}
|
|
|
|
// 2D version
|
|
void shape_pos_vel_accel_xy(const Vector2f& pos_input, const Vector2f& vel_input, const Vector2f& accel_input,
|
|
const Vector2f& pos, const Vector2f& vel, Vector2f& accel,
|
|
const float vel_correction_max, const float vel_max, const float accel_max, const float tc, const float dt)
|
|
{
|
|
if (!is_positive(tc)) {
|
|
return;
|
|
}
|
|
|
|
// Calculate time constants and limits to ensure stable operation
|
|
const float KPv = 1.0f / (CONTROL_TIME_CONSTANT_RATIO*tc);
|
|
const float accel_tc_max = accel_max*(1.0f - 1.0f/CONTROL_TIME_CONSTANT_RATIO);
|
|
|
|
// position error to be corrected
|
|
Vector2f pos_error = pos_input - pos;
|
|
|
|
// velocity to correct position
|
|
Vector2f vel_target = sqrt_controller(pos_error, KPv, accel_tc_max, dt);
|
|
|
|
// limit velocity correction to vel_correction_max
|
|
if (is_positive(vel_correction_max)) {
|
|
vel_target.limit_length(vel_correction_max);
|
|
}
|
|
|
|
// velocity correction with input velocity
|
|
vel_target = vel_target + vel_input;
|
|
|
|
shape_vel_accel_xy(vel_target, accel_input, vel, accel, vel_max, accel_max, tc, dt);
|
|
}
|
|
|
|
// proportional controller with piecewise sqrt sections to constrain second derivative
|
|
float sqrt_controller(float error, float p, float second_ord_lim, float dt)
|
|
{
|
|
float correction_rate;
|
|
if (is_negative(second_ord_lim) || is_zero(second_ord_lim)) {
|
|
// second order limit is zero or negative.
|
|
correction_rate = error * p;
|
|
} else if (is_zero(p)) {
|
|
// P term is zero but we have a second order limit.
|
|
if (is_positive(error)) {
|
|
correction_rate = safe_sqrt(2.0f * second_ord_lim * (error));
|
|
} else if (is_negative(error)) {
|
|
correction_rate = -safe_sqrt(2.0f * second_ord_lim * (-error));
|
|
} else {
|
|
correction_rate = 0.0f;
|
|
}
|
|
} else {
|
|
// Both the P and second order limit have been defined.
|
|
const float linear_dist = second_ord_lim / sq(p);
|
|
if (error > linear_dist) {
|
|
correction_rate = safe_sqrt(2.0f * second_ord_lim * (error - (linear_dist / 2.0f)));
|
|
} else if (error < -linear_dist) {
|
|
correction_rate = -safe_sqrt(2.0f * second_ord_lim * (-error - (linear_dist / 2.0f)));
|
|
} else {
|
|
correction_rate = error * p;
|
|
}
|
|
}
|
|
if (!is_zero(dt)) {
|
|
// this ensures we do not get small oscillations by over shooting the error correction in the last time step.
|
|
return constrain_float(correction_rate, -fabsf(error) / dt, fabsf(error) / dt);
|
|
} else {
|
|
return correction_rate;
|
|
}
|
|
}
|
|
|
|
// proportional controller with piecewise sqrt sections to constrain second derivative
|
|
Vector2f sqrt_controller(const Vector2f& error, float p, float second_ord_lim, float dt)
|
|
{
|
|
const float error_length = error.length();
|
|
if (!is_positive(error_length)) {
|
|
return Vector2f{};
|
|
}
|
|
|
|
const float correction_length = sqrt_controller(error_length, p, second_ord_lim, dt);
|
|
return error * (correction_length / error_length);
|
|
}
|
|
|
|
// inverse of the sqrt controller. calculates the input (aka error) to the sqrt_controller required to achieve a given output
|
|
float inv_sqrt_controller(float output, float p, float D_max)
|
|
{
|
|
if (is_positive(D_max) && is_zero(p)) {
|
|
return (output * output) / (2.0f * D_max);
|
|
}
|
|
if ((is_negative(D_max) || is_zero(D_max)) && !is_zero(p)) {
|
|
return output / p;
|
|
}
|
|
if ((is_negative(D_max) || is_zero(D_max)) && is_zero(p)) {
|
|
return 0.0f;
|
|
}
|
|
|
|
// calculate the velocity at which we switch from calculating the stopping point using a linear function to a sqrt function
|
|
const float linear_velocity = D_max / p;
|
|
|
|
if (fabsf(output) < linear_velocity) {
|
|
// if our current velocity is below the cross-over point we use a linear function
|
|
return output / p;
|
|
}
|
|
|
|
const float linear_dist = D_max / sq(p);
|
|
const float stopping_dist = (linear_dist * 0.5f) + sq(output) / (2.0f * D_max);
|
|
return is_positive(output) ? stopping_dist : -stopping_dist;
|
|
}
|
|
|
|
// calculate the stopping distance for the square root controller based deceleration path
|
|
float stopping_distance(float velocity, float p, float accel_max)
|
|
{
|
|
return inv_sqrt_controller(velocity, p, accel_max);
|
|
}
|
|
|
|
// calculate the maximum acceleration or velocity in a given direction
|
|
// based on horizontal and vertical limits.
|
|
float kinematic_limit(Vector3f direction, float max_xy, float max_z_pos, float max_z_neg)
|
|
{
|
|
if (is_zero(direction.length_squared())) {
|
|
return 0.0f;
|
|
}
|
|
|
|
max_xy = fabsf(max_xy);
|
|
max_z_pos = fabsf(max_z_pos);
|
|
max_z_neg = fabsf(max_z_neg);
|
|
|
|
direction.normalize();
|
|
const float xy_length = Vector2f{direction.x, direction.y}.length();
|
|
|
|
if (is_zero(xy_length)) {
|
|
return is_positive(direction.z) ? max_z_pos : max_z_neg;
|
|
}
|
|
|
|
if (is_zero(direction.z)) {
|
|
return max_xy;
|
|
}
|
|
|
|
const float slope = direction.z/xy_length;
|
|
if (is_positive(slope)) {
|
|
if (fabsf(slope) < max_z_pos/max_xy) {
|
|
return max_xy/xy_length;
|
|
}
|
|
return fabsf(max_z_pos/direction.z);
|
|
}
|
|
|
|
if (fabsf(slope) < max_z_neg/max_xy) {
|
|
return max_xy/xy_length;
|
|
}
|
|
return fabsf(max_z_neg/direction.z);
|
|
}
|