lol by damgaardmr · Pull Request #1 · mindThomas/Kugle-Embedded

damgaardmr · 2019-03-22T08:39:13Z

No description provided.

Linker file changed to accommodate for heap in D1 memory space which needs a pause after boot before stabilization and usage. FreeRTOS configured to support MCUXpresso FreeRTOS plugin for Eclipse for debugging.

SMBus driver prepared but not tested

RX messages are stored in message queue in USBCDC object. Library functionality is not finished though - needs functions for reading messages and transmitting, preferably queue based

…g serial characters in a similar fashion to the Arduino Serial.write and Serial.read

…loped and tested

Interrupt feature currently under test and not fully working yet

…created but estimator library development is not complete yet

Added comments for Sliding mode controller

Notice that LQR controller takes in angular velocity in body frame Notice that MATLAB Coder generated Sliding mode controller does not take in angular velocity reference

Can not compile due to missing ARM Math library during linking

…ted by the Cortex-M7 device

Verified compilation on Linux

Fixed issue with Hard Fault due to stack overflow Tested with 200 Hz duplex communication

This allows debugging messages to be sent to the PC with a simple call to "Debug::print(...)" or "Debug::printf(...)"

Verified functionality

DFU bootloader can be used to update the firmware over USB

Accelerometer calibration and gyroscope bias + alignment calibration can be carried out individually

First time the velocity controller is well working for Kugle V1

Tested sliding mode without Equivalent control and higher gains with promising results Balance is still wobbly and oscillating Allowed independent gains for the three sliding surfaces (K, eta and epsilon)

LQR controller seems to give similar performance to the Sliding mode controller ???! OBS! Velocity estimator is still not concluded functional.

1. Using q_dot sliding manifold with factor 2 error 2. Using updated/improved QEKF but in non-sensor driven configuration, with bias estimation enabled but without acceleration detection enabled 3. Using LPF for velocity estimation - thus NOT using velocity estimator 4. Using Euler-based PI velocity controller but with omega_references disabled

Q_dot sliding mode had an error in the inverse matrix (factor of two). This required the gains to be modified. However if Qdot in equivalent control is included the system becomes very sensitive and prone to noise. These gains (and without Qdot in equivalent control) works well.

Sliding mode controller gains has been greatly reduced to make the controller slower LQR Velocity controller implemented. Automatic integral start up implemented allowing the user to move and position the robot after startup.

Velocity estimation works very well with this updated Extended Kalman filter using a combination of accelerometer measurements for acceleration estimation and encoder measurements for velocity estimation/correction

- SMBus communication

mindThomas

I have now reviewed the code. Please see my comments at each file/line.
Note also that I did not find any modifications to the Include directories. Didn't you change which directories to include in the Eclipse (System Workbench) project settings such that the Battery library etc. was included?

mindThomas · 2019-03-22T08:46:10Z

KugleFirmware/Src/main.c


  /* Create the main thread which creates objects and spawns the rest of the threads */
-  xTaskCreate(MainTask, "mainTask", 1024, (void*) NULL, MAIN_TASK_PRIORITY, &mainTaskHandle);
+  xTaskCreate(MainTask, "mainTask", 1024, (void*) NULL, MAIN_TASK_PRIORITY, &mainTaskHandle); // udkommenter igen


This line should not be committed

Is it the comment that you want removed? The rest is the same

Yes, the comment should not be committed.

mindThomas · 2019-03-22T08:48:35Z

KugleFirmware/Src/MainTask.cpp

-	Battery * battery1 = new Battery();
-	Battery * battery2 = new Battery();
-	PWM * powerLED = new PWM(PWM::TIMER17, PWM::CH1, POWER_LED_PWM_FREQUENCY, POWER_LED_PWM_RANGE);
-	PowerManagement * pm = new PowerManagement(*enable19V, *enable5V, *battery1, *battery2, *powerLED, POWER_MANAGEMENT_PRIORITY);


Power management initialization can not be moved down since the 12V power should be turned on as quickly as possible to avoid the onboard computer turning off if the embedded processor is reset (eg. during debugging or just as a general reset)

However I need the LSPC and timer object in the powerManagment object, so these objects have to be initialized before powerManagment?

mindThomas · 2019-03-22T08:50:34Z

KugleFirmware/Libraries/Modules/PowerManagement/PowerManagement.cpp

+
+	if(batteryAssamblyChargePercentage<BATTERY_LOW_PERCENTAGE){
+		batteryLOW = true;
+		Debug::printf("battery low /n");


Do you want to leave this debug line in?
If so, please note that the "new line" character is '\n' and not '/n'

I think it makes sense to leave it for now, since the power LED is not yet that visible :)

OK, it makes sense. In that case please modify the new line character.

mindThomas · 2019-03-22T08:52:25Z

KugleFirmware/Libraries/Modules/PowerManagement/PowerManagement.cpp

 	_powerMode = POWERMODE_OFF;
 	Enable(true, false); // start with 19V turned on and 5V off

+	osDelay(2000);


You should not have a 2 second delay in a constructor. In that case move the below from this constructor into the Power Management thread "PowerManagement::PowerManagementThread" as an initialization part

mindThomas · 2019-03-22T08:56:25Z

KugleFirmware/Libraries/Modules/Estimators/VelocityEKF/MATLABCoder/VelocityEstimator2.h

@@ -0,0 +1,34 @@
+//


This file is wrongly committed! It should be removed from the commit and pull request.

mindThomas · 2019-03-22T08:58:40Z

KugleFirmware/Libraries/Devices/BatteryBoard/BatteryBoard.h

+ * Thomas Jespersen, TKJ Electronics
+ * Web      :  http://www.tkjelectronics.dk
+ * e-mail   :  thomasj@tkjelectronics.dk
+ * ------------------------------------------


Since you have made this you can consider to add some extra lines here in the top to indicate that you are the author of this part

mindThomas · 2019-03-22T08:59:02Z

KugleFirmware/Libraries/Devices/BatteryBoard/BatteryBoard.cpp

+ * Web      :  http://www.tkjelectronics.dk
+ * e-mail   :  thomasj@tkjelectronics.dk
+ * ------------------------------------------
+ */


Since you have made this you can consider to add some extra lines here in the top to indicate that you are the author of this part

mindThomas · 2019-03-22T09:01:37Z

KugleFirmware/Libraries/Devices/BatteryBoard/BatteryBoard.cpp

+
+	while(1){ // loop until the device becomes available and the configuration can be changed
+		if(thisObjPtr->SetCurrentLimitEEPROM()){ // stop task
+			if (thisObjPtr->batteryBoardEepromTaskHandle){ // free memory associated with the task


Isn't the comment misleading - you are not freeing memory related to the task
But I see how you suspend and resume. However there is almost no overhead in creating and destroying threads. So you could also just destroy the thread completely similar to what is done here: https://github.com/mindThomas/Kugle-Embedded/blob/master/KugleFirmware/Libraries/Applications/BalanceController/BalanceController.cpp#L962

mindThomas · 2019-03-22T09:02:53Z

KugleFirmware/KugleFirmware Debug.cfg


 # STlink Debug clock frequency
-set CLOCK_FREQ 4000
+set CLOCK_FREQ 8000


Don't commit this change. Debug frequency has been lowered purposely

mindThomas · 2019-03-22T09:03:36Z

KugleFirmware/KugleFirmware.xml

@@ -1,9 +1,19 @@
 <?xml version="1.0" encoding="UTF-8"?>


Do not commit this file change. Leave out these changes to the Eclipse configuration files.

mindThomas and others added 30 commits December 13, 2018 16:24

Initial commit

bc3cf70

Initial commit with UART driver tested and working

43389b7

SPI and I2C drivers tested against MPU9250 and confirmed working.

723f8ec

Linker file changed to accommodate for heap in D1 memory space which needs a pause after boot before stabilization and usage. FreeRTOS configured to support MCUXpresso FreeRTOS plugin for Eclipse for debugging.

PWM, Encoder, High resolution timer and IO driver implemented and tested

56043e9

SMBus driver prepared but not tested

Quadrature knob library implemented and tested

9b214aa

USB CDC functionality moved into library and has tested.

448f076

RX messages are stored in message queue in USBCDC object. Library functionality is not finished though - needs functions for reading messages and transmitting, preferably queue based

USB CDC library updated to include functions for writing and receivin…

2987d07

…g serial characters in a similar fashion to the Arduino Serial.write and Serial.read

LSPC library included and Socket port for using USB CDC has been deve…

18253e4

…loped and tested

ADC library implemented in polling configuration

ef9863f

MPU-9250 library added with support for both SPI and I2C

1beb391

Interrupt feature currently under test and not fully working yet

ESCON library implemented but not tested on hardware

f54eb2e

ESCON library tested on real hardware and confirmed working

20a3650

Controller code merged and put into classes - not tested yet

dc31fdf

Prepared application classes

1fa4548

Estimator code merged into project and library placeholders has been …

2fbf487

…created but estimator library development is not complete yet

Added LCD library but not tested yet

f758eec

Added comments for Sliding mode controller

Updated sliding mode controller to include angular velocity reference

1bc794e

Added comments to controller libraries

3af284e

Notice that LQR controller takes in angular velocity in body frame Notice that MATLAB Coder generated Sliding mode controller does not take in angular velocity reference

Estimator libraries finished but not tested yet

6af0ab2

Attitude controller application implemented (but not tested)

a641d4b

Can not compile due to missing ARM Math library during linking

Included ARM Math and DSP support to use the SIMD instructions suppor…

988c1d9

…ted by the Cortex-M7 device

Fixed several compiler warnings

ba6d095

Verified compilation on Linux

Tested and verified LSPC communication with PC

494da02

Fixed issue with Hard Fault due to stack overflow Tested with 200 Hz duplex communication

Debugging library updated to support printf and linked with LSPC (USB)

8500db6

This allows debugging messages to be sent to the PC with a simple call to "Debug::print(...)" or "Debug::printf(...)"

Modified maximum LSPC package length to 250 bytes to be within bounds

53a9cbc

Implemented unit tests for Sliding mode, LQR and QEKF

af03f24

Verified functionality

Updated copyright headers

c361ca3

Updated copyright headers

4e2c2c9

Updated copyright headers

ef1e21c

"Enter Bootloader" function tested

529a15d

DFU bootloader can be used to update the firmware over USB

mindThomas and others added 29 commits February 26, 2019 08:04

Implemented seperated IMU calibration procedure

ef3d183

Accelerometer calibration and gyroscope bias + alignment calibration can be carried out individually

Improved QEKF and Velocity estimator + Working velocity control for V1

83f8adb

First time the velocity controller is well working for Kugle V1

Velocity controller working but wobbly - note no equivalent control used

47e3731

Tested sliding mode without Equivalent control and higher gains with promising results Balance is still wobbly and oscillating Allowed independent gains for the three sliding surfaces (K, eta and epsilon)

Velocity controller working much better by using ball velocity

7e5c83b

Velocity controller tuned for Joystick control (but is still wobbly)

6fc3c68

Improved velocity controller performance by adding rate limiter

52e327c

Wheel slip detection and automatic ramp down (to reduce oscillations)

39a5028

Quaternion and Velocity control is well working + heading init at start

c65f83f

Improved wheel slip detection and velocity controller

3a094f8

LQR controller corrected and well working (similar performance to SM)

398e1ef

LQR controller seems to give similar performance to the Sliding mode controller ???! OBS! Velocity estimator is still not concluded functional.

Fixed error in 2D rotation function used by velocity controller

b6be587

Velocity estimator tested and working - but not fine tuned

3bb306d

Equivalent control only disabled around x and y axis

e22a11b

Merge branch 'SlidingModeFix' into stable

afd48b6

Merge branch 'stable'

bda54ff

Added parameters for Velocity sliding mode (see other branch)

6790c38

Accelerometer+Encoder-based Velocity estimator implemented and tested

70417ef

Velocity estimation works very well with this updated Extended Kalman filter using a combination of accelerometer measurements for acceleration estimation and encoder measurements for velocity estimation/correction

Switched to MIT License

2515cea

Merge branch 'master' into stable

eb8834d

Made powermanagment

c648ce7

- SMBus communication

comment out testBench

d51500e

Changed LSPC powermanagment msg id

3e6b1dd

Auto stash before merge of "master" and "origin/master"

51fc12c

Deleted core file

addf453

lol

5d20220

lo

6ad54da

mindThomas requested changes Mar 22, 2019

View reviewed changes

Conversation

damgaardmr commented Mar 22, 2019

Uh oh!

mindThomas left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants