klipper/klippy/toolhead.py
Kevin O'Connor 71256f9456 toolhead: Flush lookahead buffer by time
Use a minimum time window as a heuristic for determining when to try
to lazily flush the lookahead buffer.  In the common case this will
result in more moves processed for each flush and thus reduce the
overall cost of the lookahead processing.

Signed-off-by: Kevin O'Connor <kevin@koconnor.net>
2017-02-12 17:20:40 -05:00

344 lines
15 KiB
Python

# Code for coordinating events on the printer toolhead
#
# Copyright (C) 2016 Kevin O'Connor <kevin@koconnor.net>
#
# This file may be distributed under the terms of the GNU GPLv3 license.
import math, logging
import cartesian, delta, extruder
# Common suffixes: _d is distance (in mm), _v is velocity (in
# mm/second), _v2 is velocity squared (mm^2/s^2), _t is time (in
# seconds), _r is ratio (scalar between 0.0 and 1.0)
# Class to track each move request
class Move:
def __init__(self, toolhead, start_pos, end_pos, speed):
self.toolhead = toolhead
self.start_pos = tuple(start_pos)
self.end_pos = tuple(end_pos)
self.accel = toolhead.max_accel
self.is_kinematic_move = True
self.axes_d = axes_d = [end_pos[i] - start_pos[i] for i in (0, 1, 2, 3)]
self.move_d = move_d = math.sqrt(sum([d*d for d in axes_d[:3]]))
if not move_d:
# Extrude only move
self.move_d = move_d = abs(axes_d[3])
self.is_kinematic_move = False
self.min_move_t = move_d / speed
# Junction speeds are tracked in velocity squared. The
# delta_v2 is the maximum amount of this squared-velocity that
# can change in this move.
self.max_start_v2 = 0.
self.max_cruise_v2 = speed**2
self.delta_v2 = 2.0 * move_d * self.accel
self.max_smoothed_v2 = 0.
self.smooth_delta_v2 = 2.0 * move_d * toolhead.max_accel_to_decel
def limit_speed(self, speed, accel):
speed2 = speed**2
if speed2 < self.max_cruise_v2:
self.max_cruise_v2 = speed2
self.min_move_t = self.move_d / speed
self.accel = min(self.accel, accel)
self.delta_v2 = 2.0 * self.move_d * self.accel
self.smooth_delta_v2 = min(self.smooth_delta_v2, self.delta_v2)
def calc_junction(self, prev_move):
axes_d = self.axes_d
prev_axes_d = prev_move.axes_d
if (axes_d[2] or prev_axes_d[2] or self.accel != prev_move.accel
or not self.is_kinematic_move or not prev_move.is_kinematic_move):
return
# Allow extruder to calculate its maximum junction
extruder_v2 = self.toolhead.extruder.calc_junction(prev_move, self)
# Find max velocity using approximated centripetal velocity as
# described at:
# https://onehossshay.wordpress.com/2011/09/24/improving_grbl_cornering_algorithm/
junction_cos_theta = -((axes_d[0] * prev_axes_d[0]
+ axes_d[1] * prev_axes_d[1])
/ (self.move_d * prev_move.move_d))
if junction_cos_theta > 0.999999:
return
junction_cos_theta = max(junction_cos_theta, -0.999999)
sin_theta_d2 = math.sqrt(0.5*(1.0-junction_cos_theta))
R = self.toolhead.junction_deviation * sin_theta_d2 / (1. - sin_theta_d2)
self.max_start_v2 = min(
R * self.accel, self.max_cruise_v2, prev_move.max_cruise_v2
, extruder_v2, prev_move.max_start_v2 + prev_move.delta_v2)
self.max_smoothed_v2 = min(
self.max_start_v2
, prev_move.max_smoothed_v2 + prev_move.smooth_delta_v2)
def set_junction(self, start_v2, cruise_v2, end_v2):
# Determine accel, cruise, and decel portions of the move distance
inv_delta_v2 = 1. / self.delta_v2
self.accel_r = accel_r = (cruise_v2 - start_v2) * inv_delta_v2
self.decel_r = decel_r = (cruise_v2 - end_v2) * inv_delta_v2
self.cruise_r = cruise_r = 1. - accel_r - decel_r
# Determine move velocities
self.start_v = start_v = math.sqrt(start_v2)
self.cruise_v = cruise_v = math.sqrt(cruise_v2)
self.end_v = end_v = math.sqrt(end_v2)
# Determine time spent in each portion of move (time is the
# distance divided by average velocity)
self.accel_t = accel_r * self.move_d / ((start_v + cruise_v) * 0.5)
self.cruise_t = cruise_r * self.move_d / cruise_v
self.decel_t = decel_r * self.move_d / ((end_v + cruise_v) * 0.5)
def move(self):
# Generate step times for the move
next_move_time = self.toolhead.get_next_move_time()
if self.is_kinematic_move:
self.toolhead.kin.move(next_move_time, self)
if self.axes_d[3]:
self.toolhead.extruder.move(next_move_time, self)
self.toolhead.update_move_time(
self.accel_t + self.cruise_t + self.decel_t)
LOOKAHEAD_FLUSH_TIME = 0.250
# Class to track a list of pending move requests and to facilitate
# "look-ahead" across moves to reduce acceleration between moves.
class MoveQueue:
def __init__(self, extruder_lookahead):
self.extruder_lookahead = extruder_lookahead
self.queue = []
self.leftover = 0
self.next_start_v2 = 0.
self.junction_flush = LOOKAHEAD_FLUSH_TIME
def reset(self):
del self.queue[:]
self.leftover = 0
self.next_start_v2 = 0.
self.junction_flush = LOOKAHEAD_FLUSH_TIME
def flush(self, lazy=False):
self.junction_flush = LOOKAHEAD_FLUSH_TIME
update_flush_count = lazy
queue = self.queue
flush_count = len(queue)
# Traverse queue from last to first move and determine maximum
# junction speed assuming the robot comes to a complete stop
# after the last move.
delayed = []
next_end_v2 = next_smoothed_v2 = peak_cruise_v2 = 0.
for i in range(flush_count-1, self.leftover-1, -1):
move = queue[i]
reachable_start_v2 = next_end_v2 + move.delta_v2
start_v2 = min(move.max_start_v2, reachable_start_v2)
reachable_smoothed_v2 = next_smoothed_v2 + move.smooth_delta_v2
smoothed_v2 = min(move.max_smoothed_v2, reachable_smoothed_v2)
if smoothed_v2 < reachable_smoothed_v2:
# It's possible for this move to accelerate
if (smoothed_v2 + move.smooth_delta_v2 > next_smoothed_v2
or next_smoothed_v2 >= peak_cruise_v2):
# This move can both accelerate and decelerate
if update_flush_count and peak_cruise_v2:
flush_count = i
update_flush_count = False
peak_cruise_v2 = min(move.max_cruise_v2, (
smoothed_v2 + reachable_smoothed_v2) * .5)
if delayed:
# Propagate peak_cruise_v2 to any delayed moves
for m, ms_v2, me_v2 in delayed:
mc_v2 = min(peak_cruise_v2, ms_v2)
m.set_junction(min(ms_v2, mc_v2), mc_v2
, min(me_v2, mc_v2))
del delayed[:]
cruise_v2 = min((start_v2 + reachable_start_v2) * .5
, move.max_cruise_v2, peak_cruise_v2)
if not update_flush_count and i < flush_count:
move.set_junction(min(start_v2, cruise_v2), cruise_v2
, min(next_end_v2, cruise_v2))
elif not update_flush_count:
# Delay calculating this move until peak_cruise_v2 is known
delayed.append((move, start_v2, next_end_v2))
next_end_v2 = start_v2
next_smoothed_v2 = smoothed_v2
if update_flush_count:
return
# Allow extruder to do its lookahead
move_count = self.extruder_lookahead(queue, flush_count, lazy)
# Generate step times for all moves ready to be flushed
for move in queue[:move_count]:
move.move()
# Remove processed moves from the queue
self.leftover = flush_count - move_count
del queue[:move_count]
def add_move(self, move):
self.queue.append(move)
if len(self.queue) == 1:
return
move.calc_junction(self.queue[-2])
self.junction_flush -= move.min_move_t
if self.junction_flush <= 0.:
# There are enough queued moves to return to zero velocity
# from the first move's maximum possible velocity, so at
# least one move can be flushed.
self.flush(lazy=True)
STALL_TIME = 0.100
# Main code to track events (and their timing) on the printer toolhead
class ToolHead:
def __init__(self, printer, config):
self.printer = printer
self.reactor = printer.reactor
self.extruder = printer.objects.get('extruder')
if self.extruder is None:
self.extruder = extruder.DummyExtruder()
kintypes = {'cartesian': cartesian.CartKinematics,
'delta': delta.DeltaKinematics}
self.kin = config.getchoice('kinematics', kintypes)(printer, config)
self.max_speed = config.getfloat('max_velocity')
self.max_accel = config.getfloat('max_accel')
self.max_accel_to_decel = config.getfloat('max_accel_to_decel'
, self.max_accel * 0.5)
self.junction_deviation = config.getfloat('junction_deviation', 0.02)
self.move_queue = MoveQueue(self.extruder.lookahead)
self.commanded_pos = [0., 0., 0., 0.]
# Print time tracking
self.buffer_time_high = config.getfloat('buffer_time_high', 5.000)
self.buffer_time_low = config.getfloat('buffer_time_low', 0.150)
self.move_flush_time = config.getfloat('move_flush_time', 0.050)
self.print_time = 0.
self.need_check_stall = -1.
self.print_time_stall = 0
self.flush_timer = self.reactor.register_timer(self._flush_handler)
# Motor off tracking
self.motor_off_time = config.getfloat('motor_off_time', 60.000)
self.motor_off_timer = self.reactor.register_timer(
self._motor_off_handler)
def build_config(self):
# Determine the maximum velocity a cartesian axis could have
# before cornering. The 8. was determined experimentally.
xy_halt = math.sqrt(8. * self.junction_deviation * self.max_accel)
self.kin.set_max_jerk(xy_halt, self.max_speed, self.max_accel)
self.extruder.set_max_jerk(xy_halt, self.max_speed, self.max_accel)
self.kin.build_config()
# Print time tracking
def update_move_time(self, movetime):
self.print_time += movetime
flush_to_time = self.print_time - self.move_flush_time
self.printer.mcu.flush_moves(flush_to_time)
def get_next_move_time(self):
if not self.print_time:
self.print_time = self.buffer_time_low + STALL_TIME
curtime = self.reactor.monotonic()
self.printer.mcu.set_print_start_time(curtime)
self.reactor.update_timer(self.flush_timer, self.reactor.NOW)
self._reset_motor_off()
return self.print_time
def get_last_move_time(self):
self.move_queue.flush()
return self.get_next_move_time()
def reset_print_time(self):
self.move_queue.flush()
self.printer.mcu.flush_moves(self.print_time)
self.print_time = 0.
self.need_check_stall = -1.
self._reset_motor_off()
def _check_stall(self):
eventtime = self.reactor.monotonic()
if not self.print_time:
# Building initial queue - make sure to flush on idle input
self.reactor.update_timer(self.flush_timer, eventtime + 0.100)
return
# Check if there are lots of queued moves and stall if so
while 1:
buffer_time = self.printer.mcu.get_print_buffer_time(
eventtime, self.print_time)
stall_time = buffer_time - self.buffer_time_high
if stall_time <= 0.:
break
eventtime = self.reactor.pause(eventtime + stall_time)
if not self.print_time:
return
self.need_check_stall = self.print_time - stall_time + 0.100
def _flush_handler(self, eventtime):
try:
if not self.print_time:
# Input idled before filling lookahead queue - flush it
self.move_queue.flush()
if not self.print_time:
return self.reactor.NEVER
print_time = self.print_time
buffer_time = self.printer.mcu.get_print_buffer_time(
eventtime, print_time)
if buffer_time > self.buffer_time_low:
# Running normally - reschedule check
return eventtime + buffer_time - self.buffer_time_low
# Under ran low buffer mark - flush lookahead queue
self.move_queue.flush()
if print_time != self.print_time:
self.print_time_stall += 1
self.dwell(self.buffer_time_low + STALL_TIME)
return self.reactor.NOW
self.reset_print_time()
except:
logging.exception("Exception in flush_handler")
self.force_shutdown()
return self.reactor.NEVER
# Motor off timer
def _reset_motor_off(self):
if not self.print_time:
waketime = self.reactor.monotonic() + self.motor_off_time
else:
waketime = self.reactor.NEVER
self.reactor.update_timer(self.motor_off_timer, waketime)
def _motor_off_handler(self, eventtime):
try:
self.motor_off()
self.reset_print_time()
except:
logging.exception("Exception in motor_off_handler")
self.force_shutdown()
return self.reactor.NEVER
# Movement commands
def get_position(self):
return list(self.commanded_pos)
def set_position(self, newpos):
self.move_queue.flush()
self.commanded_pos[:] = newpos
self.kin.set_position(newpos)
def move(self, newpos, speed):
speed = min(speed, self.max_speed)
move = Move(self, self.commanded_pos, newpos, speed)
if not move.move_d:
return
if move.is_kinematic_move:
self.kin.check_move(move)
if move.axes_d[3]:
self.extruder.check_move(move)
self.commanded_pos[:] = newpos
self.move_queue.add_move(move)
if self.print_time > self.need_check_stall:
self._check_stall()
def home(self, homing_state):
self.kin.home(homing_state)
def dwell(self, delay):
self.get_last_move_time()
self.update_move_time(delay)
self._check_stall()
def motor_off(self):
self.dwell(STALL_TIME)
last_move_time = self.get_last_move_time()
self.kin.motor_off(last_move_time)
self.extruder.motor_off(last_move_time)
self.dwell(STALL_TIME)
logging.debug('; Max time of %f' % (last_move_time,))
def wait_moves(self):
self.move_queue.flush()
eventtime = self.reactor.monotonic()
while self.print_time:
eventtime = self.reactor.pause(eventtime + 0.100)
def query_endstops(self):
last_move_time = self.get_last_move_time()
return self.kin.query_endstops(last_move_time)
# Misc commands
def stats(self, eventtime):
buffer_time = 0.
if self.print_time:
buffer_time = self.printer.mcu.get_print_buffer_time(
eventtime, self.print_time)
return "print_time=%.3f buffer_time=%.3f print_time_stall=%d" % (
self.print_time, buffer_time, self.print_time_stall)
def force_shutdown(self):
self.printer.mcu.force_shutdown()
self.move_queue.reset()
self.reset_print_time()