OHMC2019 Software instructions - Revision history

Marcmerlin: /* Neopixels */

2019-01-24T04:14:16Z

Neopixels

@@ Line 237: / Line 237: @@
 Then, you need to run as root for neopixels to work:
 <pre>
-sudo /home/pi/env/bin/python manage.py drive --model=/home/pi/ohmc_car/models/large_track.hdf5
+sudo /home/pi/env/bin/python manage.py drive
 </pre>

Marcmerlin: Added neopixel info

2019-01-21T09:06:26Z

Added neopixel info

← Older revision		Revision as of 09:06, 21 January 2019
Line 199:		Line 199:

	It is recommended to just start with Local Angle ... and control the throttle manually with the "i" key (faster) and "k" key (slower).		It is recommended to just start with Local Angle ... and control the throttle manually with the "i" key (faster) and "k" key (slower).


			== Neopixels ==

			On the car, do this: pip install rpi_ws281x adafruit-circuitpython-neopixel --trusted-host pypi.org --trusted-host files.pythonhosted.org --trusted-host www.piwheels.org

			The details come from here: https://learn.adafruit.com/neopixels-on-raspberry-pi/overview

			<pre>
			(env) pi@ohmc_24:~/ohmc_car$ diff -u manage.py.orig manage.py
			--- manage.py.orig 2018-06-11 08:14:34.079999106 +0000
			+++ manage.py 2018-06-11 08:37:05.959999560 +0000
			@@ -122,6 +122,9 @@
			tub = TubWriter(path=cfg.TUB_PATH, inputs=inputs, types=types)
			V.add(tub, inputs=inputs, run_condition='recording')

			+ pixels[0] = (32, 64, 128)
			+ pixels[1] = (0, 255, 0)
			+
			# run the vehicle
			V.start(rate_hz=cfg.DRIVE_LOOP_HZ,
			max_loop_count=cfg.MAX_LOOPS)
			@@ -169,6 +172,13 @@
			cfg = dk.load_config()

			if args['drive']:
			+ # Neopixels
			+ # https://learn.adafruit.com/neopixels-on-raspberry-pi/python-usage
			+ import board
			+ import neopixel
			+ pixels = neopixel.NeoPixel(board.D18, 2)
			+ pixels[0] = (0, 0, 0)
			+ pixels[1] = (0, 64, 0)
			drive(cfg, model_path=args['--model'], use_chaos=args['--chaos'])
			</pre>

			Then, you need to run as root for neopixels to work:
			<pre>
			sudo /home/pi/env/bin/python manage.py drive --model=/home/pi/ohmc_car/models/large_track.hdf5
			</pre>

Marcmerlin: improve security of data retrieval from car

2019-01-21T08:50:39Z

improve security of data retrieval from car

← Older revision		Revision as of 08:50, 21 January 2019
Line 144:		Line 144:
	You will need to type this command just once on your DonkeyCar (Raspberry Pi)to provide a directory on your laptop for your training data ...		You will need to type this command just once on your DonkeyCar (Raspberry Pi)to provide a directory on your laptop for your training data ...

	* ~~scp~~ -~~pr ohmc_car $USERNAME~~@~~$HOSTNAME~~:		* rsync -av pi@<car_ip>:ohmc_car .

	When finished acquisition, then transfer the data from the DonkeyCar (Raspberry Pi) to your laptop / desktop for training the Neural Network.		When finished acquisition, then transfer the data from the DonkeyCar (Raspberry Pi) to your laptop / desktop for training the Neural Network.

	* ~~scp~~ -~~pr tub $USERNAME~~@~~$HOSTNAME~~:ohmc_car/tub_$DATE		* rsync -av pi@<car_ip>:ohmc_car/tub ohmc_car/tub_$DATE

	== Training the Neural Network ==		== Training the Neural Network ==

Marcmerlin: work around mkvirtualenv not being there

2019-01-21T07:30:25Z

work around mkvirtualenv not being there

← Older revision		Revision as of 07:30, 21 January 2019
Line 14:		Line 14:
	(if you get errors, you can try (re-) installing pip: python -m pip install --upgrade pip )		(if you get errors, you can try (re-) installing pip: python -m pip install --upgrade pip )

	* cd $~~HOME~~		On debian, if virtualenv isn't there, try this instead
			* virtualenv donkeycar -p python3
			* cd donkeycar
			* export PATH=`pwd`/bin:$PATH
			* pip install tensorflow==1.8.0

			Then proceed:

	* git clone https://github.com/autorope/donkeycar		* git clone https://github.com/autorope/donkeycar
	* pip install -e ./donkeycar		* pip install -e ./donkeycar

Marcmerlin: Added info on how to deal with rPi reboots

2019-01-21T06:11:42Z

Added info on how to deal with rPi reboots

← Older revision		Revision as of 06:11, 21 January 2019
Line 98:		Line 98:

	The mobile web browser, the user interface allows you to drive by tilting the phone left-right for steering and forwards-backwards for throttle. For safety, you must press the [Start Vehicle] / [Stop Vehicle] toggle button to enable control.		The mobile web browser, the user interface allows you to drive by tilting the phone left-right for steering and forwards-backwards for throttle. For safety, you must press the [Start Vehicle] / [Stop Vehicle] toggle button to enable control.

			== rPi reboots and Auto-starting donkeycar without logging in ==

			If your rPi reboots when you're driving, check the following things:
			# did you swap the servo wires between steering and power. Turn the ESC off, unplug one servo wire, turn the ESC back on (little toggle switch), and test steering. Make sure that steering does not activate the wheel motors.
			# is your battery low? are you accelerating too quickly?
			Because the battery has a high internal resistance, if you demand a lot of power too quickly, the voltage on the battery drops enough that the rPi gets below its critical voltage (when you add the voltage drop from the voltage converter), and the rPi reboots.

			To restart donkey server automatically, do the following:
			* Add this to /etc/rc.local:
			su - pi -c 'cd ~/ohmc_car; python manage.py drive &>/tmp/out' &
			* move 'source ~/env/bin/activate' from ~/.bashrc to ~/.profile

	== Steering and throttle calibration ==		== Steering and throttle calibration ==

Geekscape: /* Self-driving the car */

2019-01-21T00:31:42Z

Self-driving the car

← Older revision		Revision as of 00:31, 21 January 2019
Line 163:		Line 163:
	* scp $USERNAME@$HOSTNAME:ohmc_car/models/model_$DATE.hdf5 models		* scp $USERNAME@$HOSTNAME:ohmc_car/models/model_$DATE.hdf5 models

	== Self-driving the ~~car~~ ==		== Self-driving ... the ultimate goal ==

			Once your trained model has been copied back onto the DonkeyCar (Raspberry Pi), your car can be self-driven as follow ...

			* python manage.py drive --model ~/ohmc_car/models/models/model_$DATE.hdf5

			This works similar to the manual driving mode with the addition of a trained model that can either ...

			1) User: Manual control of both steering and throttle

			2) Local Angle: Automatically control the steering angle

			3) Local pilot: Automatically control both the steering angle and throttle amount

			The web browser provides a drop-down menu to select between these options.

			It is recommended to just start with Local Angle ... and control the throttle manually with the "i" key (faster) and "k" key (slower).

Geekscape: /* Training the Neural Network */

2019-01-21T00:22:13Z

Training the Neural Network

← Older revision		Revision as of 00:22, 21 January 2019
Line 137:		Line 137:
	[http://docs.donkeycar.com/guide/train_autopilot Extensive DonkeyCar documentation]		[http://docs.donkeycar.com/guide/train_autopilot Extensive DonkeyCar documentation]

	~~Use~~ these commands to train the Neural Network ...		Run these commands on your laptop / desktop to train the Neural Network ...

	* workon donkeycar # For those who have set-up a virtualenv		* workon donkeycar # For those who have set-up a virtualenv

Geekscape: /* Data acquisition for the Neural Network */

2019-01-21T00:20:39Z

Data acquisition for the Neural Network

← Older revision		Revision as of 00:20, 21 January 2019
Line 125:		Line 125:
	You will need to type this command just once on your DonkeyCar (Raspberry Pi)to provide a directory on your laptop for your training data ...		You will need to type this command just once on your DonkeyCar (Raspberry Pi)to provide a directory on your laptop for your training data ...

	* scp -pr ohmc_car $USERNAME@$HOSTNAME:~~$HOME~~		* scp -pr ohmc_car $USERNAME@$HOSTNAME:

	When finished acquisition, then transfer the data from the DonkeyCar (Raspberry Pi) to your laptop / desktop for training the Neural Network.		When finished acquisition, then transfer the data from the DonkeyCar (Raspberry Pi) to your laptop / desktop for training the Neural Network.

	* scp -pr tub $USERNAME@$HOSTNAME:~~$HOME/~~ohmc_car/tub_$DATE		* scp -pr tub $USERNAME@$HOSTNAME:ohmc_car/tub_$DATE

	== Training the Neural Network ==		== Training the Neural Network ==

Geekscape: /* Training the Neural Network */

2019-01-21T00:20:08Z

Training the Neural Network

← Older revision		Revision as of 00:20, 21 January 2019
Line 137:		Line 137:
	[http://docs.donkeycar.com/guide/train_autopilot Extensive DonkeyCar documentation]		[http://docs.donkeycar.com/guide/train_autopilot Extensive DonkeyCar documentation]

	Use these commands ~~want you want~~ to train the Neural Network ...		Use these commands to train the Neural Network ...

	* workon donkeycar # For those who have set-up a virtualenv		* workon donkeycar # For those who have set-up a virtualenv
Line 154:		Line 154:
	Epoch 00001: val_loss improved from inf to 0.19699, saving model to ./models/roba0_2019-01-16c.hdf5		Epoch 00001: val_loss improved from inf to 0.19699, saving model to ./models/roba0_2019-01-16c.hdf5
	44/44 [==============================] - 38s 874ms/step - loss: 57.1887 - angle_out_loss: 29.6601 - throttle_out_loss: 84.7172 - val_loss: 0.1970 - val_angle_out_loss: 0.3230 - val_throttle_out_loss: 0.0710		44/44 [==============================] - 38s 874ms/step - loss: 57.1887 - angle_out_loss: 29.6601 - throttle_out_loss: 84.7172 - val_loss: 0.1970 - val_angle_out_loss: 0.3230 - val_throttle_out_loss: 0.0710

			On a modern laptop, each epoch will take around 30 seconds to complete. For up-to 100 epochs. Typically, you can expect around 20 to 40 epochs before the Neural Network stop learning. That is around 10 to 20 minutes of training time.

	The training command creates the Neural Network weights that represent what your DonkeyCar has "learned".		The training command creates the Neural Network weights that represent what your DonkeyCar has "learned".

			When training has completed, copy the trained model back to your DonkeyCar (Raspberry Pi)

			* scp $USERNAME@$HOSTNAME:ohmc_car/models/model_$DATE.hdf5 models

	== Self-driving the car ==		== Self-driving the car ==

Geekscape: /* Training the Neural Network */

2019-01-21T00:15:19Z

Training the Neural Network

← Older revision		Revision as of 00:15, 21 January 2019
Line 137:		Line 137:
	[http://docs.donkeycar.com/guide/train_autopilot Extensive DonkeyCar documentation]		[http://docs.donkeycar.com/guide/train_autopilot Extensive DonkeyCar documentation]

	Use these commands ~~every time~~ you want to train the Neural Network ...		Use these commands want you want to train the Neural Network ...

	* workon donkeycar # For those who have set-up a virtualenv		* workon donkeycar # For those who have set-up a virtualenv
	* cd ohmc_car		* cd ohmc_car
	* python manage.py train --tub ./~~tub~~ --model ./models/model_$DATE.hdf5		* python manage.py train --tub $HOME/ohmc_car/tub_$DATE --model ./models/model_$DATE.hdf5

			using donkey version: 2.5.7 ...
			loading config file: /Users/andyg/play/ai/roba_car/config.py
			config loaded
			tub_names ./tub_2019-01-15c
			train: 5740, validation: 1436
			steps_per_epoch 44
			Epoch 1/100
			2019-01-21 13:08:49.507048: I tensorflow/core/platform/cpu_feature_guard.cc:140] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX2 FMA
			43/44 [============================>.] - ETA: 0s - loss: 58.5130 - angle_out_loss: 30.3421 - throttle_out_loss: 86.6839
			Epoch 00001: val_loss improved from inf to 0.19699, saving model to ./models/roba0_2019-01-16c.hdf5
			44/44 [==============================] - 38s 874ms/step - loss: 57.1887 - angle_out_loss: 29.6601 - throttle_out_loss: 84.7172 - val_loss: 0.1970 - val_angle_out_loss: 0.3230 - val_throttle_out_loss: 0.0710

			The training command creates the Neural Network weights that represent what your DonkeyCar has "learned".

	== Self-driving the car ==		== Self-driving the car ==