PID (Floating Point)

PID is an algorithm that uses feedback to control an output.  It is used in many embedded applications to control temperature, motor speed, etc.  It is based on first calculating the error (set point minus present value) then adjusting the output based on three constants multiplied by the error:  proportional, integral, and differential.  This code snippet is an implementation in C using floating point variables.

/*! \details This structure holds the data to run a
 * floating point PID loop.
 */
typedef struct{
	float max /*! \brief Max manipulated value */;
	float min /*! \brief Miniumum manipulated value */;
	float e /*! \brief Error value */;
	float i /*! \brief Integrator value */;
	float kp /*! \brief Proportional constant */;
	float ki /*! \brief Integrator constant */;
	float kd /*! \brief Differential constant */;
} pid_f_t;

/*! \details This function initializes the data in a PID structure.
 *
 */
void pid_init_f(pid_f_t * ptr /*! A pointer to the PID data structure */,
		float min /*! The manipulated variable's minimum value */,
		float max /*! The manipulated variable's maximum value */){
	memset(ptr, 0, sizeof(pid_f_t));
	ptr->min = min;
	ptr->max = max;
}

/*! \details This function updates the value of the manipulated variable (MV)
 * based on the current state of the PID loop.
 */
float pid_update_f(float sp /*! The set point */,
		float pv /*! The process variable */,
		pid_f_t * ptr /*! A pointer to the PID constants */){
	float temp;
	float e;
	float p;
	float manp;
	float tmpi;
	//get the error from the last call
	e = ptr->e;
	//calculate the new error (set point - present value)
	ptr->e = sp - pv;
	//use a temp variable for the integrator
	tmpi = ptr->i + ptr->e;
	//update the manipulated process variable
	manp = ptr->kp * ptr->e + ptr->ki * tmpi + ptr->kd * (ptr->e - e);
	//the integrator is only updated if the manipulated process is within range
	//otherwise the system will likely become unstable
	if ( (manp < ptr->max) && (manp > ptr->min) ){
		ptr->i = tmpi;
	} else if ( manp > ptr->max ){
		manp = ptr->max;
	} else if ( manp < ptr->min ){
		manp = ptr->min;
	}
	return manp;
}