This is an implementation of a multiprocess egg dispenser thought and builded to support interchangable input and output modules.
| File name | Description |
|---|---|
| main.c | Mainly starts the managers |
| manager_io.c | Logical manager and middle process between manager_input.c and manager_output.c |
| manager_input.c | Input module |
| manager_output.c | Output module |
For a standard installation we do encourage to download and build the project on your own with the following
git clone https://github.com/draane/egg-robotic-arm.git
cd egg-robotic-arm
make
The process is thought to be placed as a systemd service and, for the default modules, it should not be started as root, because it is not necessary. Please do see the DebianWiki for more information
graph TD
A((main)) --> B((manager_io))
B --> E((manager_input))
B --> C((manager_output))
graph TD
A((manager_io)) --> B(spawn manager_input)
B(spawn manager_input) --> E(spawn manager_output)
E(spawn manager_input) --> C(trigger input)
C(trigger input) --START_MSG--> D(read input response)
D(read input response) --> F(trigger output)
F(trigger output) --START_MSG--> G(read output response)
G(read output response) --> C(trigger input)
graph TD
A((manager_input)) --> B(spawn childs)
B(spawn childs) --> E(waits for manager_io)
F((child)) --> G(read GPIO)
F((child)) --> I(signal reciver)
G(read GPIO) --> H(ellab and save)
H(ellab and save) --> G(read GPIO)
E(waits for manager_io)--signal-->I(signal reciver)
I(signal reciver) --result-->L(send manager_io)
L-->E
graph TD
A((manager_output)) --> B(spawn childs)
B(spawn childs) --> E(recive data from manager io)
F((child)) --> G(signal reciver)
G-->E
G --> H(write on GPIO)
H --> G
E--SIG_USR1 OR SIG_USR2-->G
In general inter process communication between the managers is achieved via a strict protocol that every module has to comply to. Failing to follow the protocol flow will result in the program closing itself with a non-zero status. The protocol itself is based upon a pipe-based full duplex communication. The creation of the pipes is the manager_io job as better explained further in this document. Every child process expects to recive the content of the START_MSG macro defined in the utils.h header file.
Note: The size of the START_MSG MUST not exceed the MAX_INFO_TO_SEND_SIZE macro but ignorance of this rule does not lead to failed compilation although will cause the child process to be impossible to trigger.
Once the child process recives the start message the response will be placed in the response pipe. The response is process dependant and it will be further discussed in the following paragraphers
Tip: All the process do follow the standard that states that the pipe 0 is the one from where to read and the pipe 1 is the one where to write but we do encourage to use the READ_PIPE and WRITE_PIPE macros because it makes the code more clean.
The manager io basically makes any communication between the others managers possible being the middle process and host of most of the logic. This process has control over the communication flow and it has the job to trigger the managers. The flow is well described in the previous section, the further will focus more on the logic and such than else.
Once the the input is recived in a char from the input manager it needs to be translated to be parsable by the output manager. Basically the char* is used as a bool array so that the first 6 char are either 'a' ( empty ) or 'b' ( present ) for the egg box sensors. The last 2 chars are the values in the warehouse.
Tip: a and b are not constants, they can be changed in utils.h. Basically they are calculated using the OFFSET_OUTPUT_MSG + ON or OFF macros.
The following is going to be splitted into two main part because as such the logic of the input manager is divided
The main part has the responsibility to set things up and eventually communicate with the manager io when needed. The main part uses pipes to send data to the father as above described. Into itself, to communicate with the children, pipes and signal are used. When the triggering message is recived the main part sends out a signal to his children and then reads the output they provide from the individual pipe that they share.
In order to make the children as indipendent as possible, given that maybe they could have some heavy interpretation of the input to do, what was build was a system relying on signals to make the polling of the results. This way, once a child has elaborated the input and saved it, it could go on and elaborate the next one without having to waste time waiting for his father to need it and ask for it. Once the signal is recived from the father the result of the previous operation is sent in the signal handler.
NOTE: The variable used to save the value is of type integer on most of the systems, because that is the implementation of that type that usually is adopted. That special type is used to ensure that the equal operation is atomic
As the manager input, the manager output is mainly splitted in two parts, the main part, that is responsible to communicate with the manager_io, and the child part, that does what it gets ordered.
The main part set things up, like spawning the childs, enabling the pins and starting the serial comunucation with the arm. After that it idles until data is recived from the manager_io. Once this happened the data gets parsed and forwarded to the child process using signals, then comunicate with the robotic arm, if installed, sending data about where and how many eggs move.
The child process idles and is limited to signal elaboration. Once a signal is recived it sets the correct GPIO to the correct status and goes back to idle.
The frontend is managed through an arduino nano that is a layer of abstraction between the GPIO, that expect a digital bit perfect signal on the pins, and the hardware that usually does need some fuzzling to have the expected behavior. The Arduino's job is to work just with the hardware so no major schematic is really needed to explain his code but the circuit is shown under.