MASSAGE mode is now useable again for now.
(no more random driving and unexpected shaking)
A full rework is still planned.
- Fix setFade method (did not change anything when not writing to nvs)
- joystick/massage:
- reverse motor direction every 2nd cycle to preving driving
- simplify modes (now only FW/REV and LEFT/RIGHT)
- use joystick angle to vary the duty cycle
- optimize delay and duty calculation (min max config)
- drop hysteresis (not much use, just much code)
- fix deceleration being overwritten in nvs at massage mode change (as accel before)
- handle deadlock when handle-loop does not give back mutex (rare but had that once -> restart)
- clear display when changing to mode-select
- clear display when switching status page
- adjust fan time (less on)
- slightly adjust accel/decel/max duty (needs testing)
Tested newly implemented menu on armchair and made several Fixes:
- reduce long delay when changing to menu modes (control handle)
- add unused line to mode select menu, also drop "click to confirm"
- fix and optimize getNextSelectableModeIndex function to get n next (recursive, offset)
- button: clear encoder queue before changing to menu mode
- fix menu immediately exiting on x5 press (missing break)
General:
- Rename control mode MENU to MENU_SETTINGS
- Add new control mode MENU_MODE_SELECT
Button menu:
- change 'long press' to: 'switch to menu mode-select'
- move 'open settings menu' to 5x press
Display:
- display_task: outsource statusScreen to local function
- display_task: also handle MENU_MODE_SELECT control mode
- show display startup message for 2600ms instead of 2000ms
Menu:
- Add functions to handle modeSelect menu to select from a list of allowed modes
(similar to settings main menu)
In any mode except MENU the encoder rotateion selects status screens
Add functions to rotate through available status screens
Aandle rotate encoder event in button menu
Briefly tested this - brake works as intended
- Add config options brakeDecel and brakePauseBefureResume
- control: update brakeStartThresholdDuty at startup
and maxDuty change for each motor
- motorctl: drop/simplify decel boost to brake
(faster deceleration at certain threshold)
When moving the joystick in other direction than driving
the deceleration gets speed up.
The boost scales with how far the stick is in other
direction up to a maximum
Some issues with this approach see notes in code
when moving the joystick in opposite direction the
deceleration gets boosted depending on how large
the opposite target duty is up to a maximum
Needs testing, switching between ROTATE mode might
cause issues
- boost when turning significantly improved driving experience
- display screensaver (timeouts, moving text, brightness)
- bugfixes
- different approach in motorctl implementing different control modes (needs testing)
Stable to use for now, but still WIP due to some bugs
needs optimizations and issues to be fixed before release
Revert change where massage commands were only generated at
joystick change, since it has to be handled
frequently in any case for motors to actually stop after certain time.
Not tested yet
sdkconfig:
- increase encoder dead-time to fix bug where encoder triggered multiple
short pressed events at one press.
E.g. instantly submitted value when entering a menu page sometimes
button:
- decrease input-timeout (long press time) to 500ms same as encoder long-press
- empty encoder queue when changing to MENU
- re-enable or increase joystick scaling to have more resolution at
slower speeds
Add mutex to fix bug motors stay on when mode-change while driving
due to race condition when handle() still executing while mode change
Change joystick scaling parameters
control:
- add mutex to handle() and changemode() to prevent race condition
- outsource handle() method from createHandleLoop()
- change joystick scaling (reduce 'more detail in slower area')
- comments, formatting
Rework joystick command generation
Fix timeout no commands received in motorctl
Successfully tested this state on actual hardware:
turning behavior is significantly improved - does not get slower when turning anymore
joystick:
- add boost of inner tire when turning
- add threshold where ratio snaps to 1
- optimize structure, logging
control:
- rename maxDuty to maxDutyStraight to be more clear
- add methods to change and get new variable RelativeBoostPer for menu item
menu:
- add new item to set maxRelativeBoost config parameter
motorctl:
- fix timeout not working:
previously when not receiving commands for 15s the duty was set to 0 for 1 handle cycle only
As initially planned starting wifi only when needed, at change to HTTP mode.
Tested this on breakout board: ESP32 is not getting hot anymore!
Also optimized timeout: only notify "forgot to turn off" when battery is below certain
threshold (e.g. disable when charger is connected over night)
main:
- dont start wifi at startup anymore
- remove double-initialization of nvs
already initialized in wifi.c, thus removed some code
- optimize comments, logging
control:
- timeout "forgot to turn off": Add battery threshold
- start/stop wifi when switching to/from HTTP mode
wifi:
- split init function to separate functions for NVS and NETIF (more clear in main)
- optimize log output at nvs init (moved from main)
- rename start/stop functions, formatting
Compared to the white-text display used for testing,
the actually used blue-text display does not support
hardware-scrolling (the text was just static)
display:
- Add code that continously scrolls screensaver text 1 character
to the right and wraps the truncating part to the start
- Add option to enable/disable hardware-scrolling if available
control:
- reset timeout on user input only
-> drop reset on changed motor duty
- add timeoutNotifyPowerStillOnMs
-> when forgotten to turn off the power the buzzer
beeps a few times every 30 minutes
- Add/fix JOYSTICK_LOG_IN_IDLE option
remove empty config.hpp
Tested new control modes on actual hardware.
Adjust new code so the modes kind of work now, as proof of concept.
Still needs major optimization and fixes though.
motorctl:
- add config option to disable logging for particular instance
- add some definitions to finetune control modes
- rework current and speed mode, so they actually kind of work
- fix TCS to not cause deadlock motors off
menu:
- add item set motorControlMode (select DUTY, CURRENT, SPEED)
- fix missing item maxDuty
speedsensor: fix return type
very experimental needs testing/debugging,
other control modes can currently be selected by editing the class definition in motorctl.hpp
menu/config
- add menu item to enable/disable traction control system
main: pass ptr to other motor to motor object
speedsensor: add method to get last update time
motorctl: handle loop:
- re-arrange some code sections
- add several methods to get current status (needed from other motor for tcs)
- add sketchy code for different control modes DUTY, CURRENT, SPEED (very basic implementation)
- add experimental code for traction control
display.cpp:
Optimize battery voltage measurement
While calibrating noticed it is very non-linear
- outsource function to scale using lookup table
- add lookup table to batvoltage measurement
inserted many values to lookup table while testing
currentsensors:
Fix current value jumping around between 0 and 0.2 on display
while standstill
- add parameter snapToZeroThreshold
Add new status screens that can be selected in menu
(status screen = display content while not in MENU mode)
- display:
- Motors screen: shows Power usage, duty, speed
- Joystick screen: shows all stick data and current control mode (similar to 'debug joystick' menu option
- menu:
- disable example menu items
- add new status screens
motorctl, currentsensor:
- add config option for inverted current sensor
- adjust loglevels
config:
- Adjust gpio pins to actual wiring (not breakout board for testing)
- Add min pulse durations for speedsensors (measurements with scope)
- Adjust config to currently mounted encoders
control, buzzer:
- adjust beeping
- Add feature for optinal delay to buzzer class:
have some pause after beeps instead of immediately
continuing with next queued sequence
Tested and debugged speed measurement on actual hardware.
Speed measurement is very stable now (reliable direction)
- fix direction inverted no effect
- fix direction detection
- remove "invalid directions" since every combination is valid
- trigger on rising edge only
- ignore too short pulses - possible noise (config option)
- ignore invalid pulse orders
- debug log count of ignored sequences
- config, chairAdjust:
- disable reset-pin of display (not connected, random pin was assigned)
same pin was used for chair adjust resulting in on state after startup
- fix chairAdjust not stopping in OFF state
- display
- add functionality to show different status screens (content when not in MENU)
- add status screen 'Speed' (to show speed and debug speedsensors)
- menu
- fix item joystick define center not working
- add item to select status screen
Major changes in motorctl and control to optimize performance
by freeing unnecessary cpu usage by motorctl task
Needs testing on actual hardware!
motorctl:
- slow down handle loop when duty is at target (wait for new command)
- create separate task for each motor
- setTarget method also accepts motorCommand directly now
control.cpp:
- redurce stress on motorctl by removing unnecessary commands
- set motors to idle at mode change only, instead of every iteration (IDLE, MENU, ADJUST)
- HTTP, JOYSTICK: only update motors when stick data actually changed
- simplify code
- add method for idling both motors
- use motorcommands directly in setTarget()
http:cpp:
- dont block control task with getData() method
- handle timeout independent of one queue event
- prevents unresponsive system for http-timeout when changing mode from HTTP
- menu:
- Add item joystick calibration wizzard
- Add item RESET (clear nvs and restart)
- Add configuration option to not show/change value. Instead only show confirm message
apply this to all items only running action
- Optimize formatting and comments
- main:
- pass nvsHandle to display task and joystick class
- joystick:
- Add methods to write and load calibration values (axis min/max adc value) from nvs
- Add method to get raw adc value
menu:
- optimize set-value page: send static content only once
so only update value -> significant performance boost
- formatting
- reset control timeout to prevent unintended exit and bugged display
- menu item: add option to show a default value (function ptr)
- add default value to adjust fading
- add beeping
motorctl:
- dont write to nvs when value is unchanged
Changes in menu for accel/decel time are now persistent after
restarts
main: initialize nvs and pass pointer to motorctl task
motorctl:
- add method to get default configured value
- add name to config -> adjust logging
- add methods to read and write msFadeAccel and msFadeDecel from nvs
Even though the object was created in STACK using alloca
apparently the method call via pointer in lambda function still
takes quite long since the encoder in MENU started to lag again
It gets initialized before main() now this fixed the lag
- outsource configuration of display and encoder from source/header file
to config.cpp and pass it to init function or task from main()
- optimize logging in several init functions
- speedsensor:
- fix startup error: initialize ISR only once
- create instances at initialization instead of first method call
- ssd1306 display library:
- modify library to pass offsetX to init function instead of using macro
While testing the new commits on actual hardware noticed
the motors always beeing on at full power.
-> This was due to uninitialized value. It had to fade down from
duty 81236182734612738 which takes forever.
-> This also fixes the old issue where left motor jumped for a
second after startup
Also tested JOYSTICK, HTTP, ADJUST_CHAIR modes and encoder+menu
While testing the ADJUST_CHAIR mode on actual hardware fixed the
following issues:
- main.cpp: Uninitialized pointer for leg/back-rest were passed to
control task thus as when a method was called the controller crashed
- chairAdjust.cpp: the state was never reset to REST_OFF when below
stick threshold
- outsource task_fans and task_buzzer from main.cpp to their source files
- use task parameters to pass necessary configs and objects
- adjust task priorities (display was too low)
Since moving all objects to heap encoder started to lag
interrupt not recognizing some single events in MENU
fast executen of motordriver setTarget caused the lag,
initialize it in STACK while still using a pointer now
