Crates.io | sac-base |
lib.rs | sac-base |
version | 0.0.7 |
source | src |
created_at | 2020-03-18 18:42:42.017739 |
updated_at | 2020-05-10 20:55:36.092228 |
description | Base crate of the sac-signal and sac-control crates |
homepage | |
repository | https://gitlab.com/signal-and-control/sac-base |
max_upload_size | |
id | 220242 |
size | 83,308 |
sac-base is the base crate of its two child crates sac-control and sac-signal. This crate is still under heavy construction and not finished.
The sac-* crates help to implement signal processing and control systems in a network based approach. Similar to to Simulink as example. The project is designed for a primary usage on embedded systems and is therefore a [no_std] crate. A generic solution such as the used network based approach will never be as fast as a custom system implementation for a specific problem. But it will allow to directly and quickly implement and change systems in code instead of designing it Matlab and exporting it to C. Generation the network and building the topology is relatively expensive. Therefore the crate does all the allocations and setup before the normal operation starts. This base crate is designed in a way that allows the Signal and the Control crate to easily extend it, and allow the operations to be interchangeable between each other. Both the sac-control and the sac-signal are under construction and are not available yet.
The base crate provides the following features:
The library is built with no_std. Therefore it can be used on embedded systems and while not tested, probably wasm aswell.
The following example, can be build by the code given below:
//
// +----+
// | |
// +--+ In +----+
// | | | +-----+
// +----+ +---+ |
// | Add +----+-------------+
// +----+ +---+ | | |
// | | | +-----+ | | +-----+
// +--+ In +----+ | +-----+ |
// | | | | Mul +------+---+
// +----+ | +-----+ | |
// | +-----+ | +-----+ |
// +----+ +--+ | | |
// | | | Add | | |
// +--+ In +----------------------+ +----+ |
// | | | | |
// +----+ +-----+ | |
// | +-----+ |
// | +-----+ |
// | | | |
// +--------------+ z^1 +--------------+
// | |
// +---- +
//
use sac_base::network::network::Network;
use sac_base::operations::math::add::Add;
use sac_base::operations::miscellaneous::buffer::Buffer;
use sac_base::operations::math::multiply::Multiply;
// Generate the network
let mut network = Network::new();
// Generate some operations
let add1 = network.add_operation(Add::new());
let mul1 = network.add_operation(Multiply::new());
let add2 = network.add_operation(Add::new());
// Add some inputs
let in1 = network.add_input(Buffer::new());
let in2 = network.add_input(Buffer::new());
let in3 = network.add_input(Buffer::new());
// Connect the operations with each other
network.add_connection(&in1, &add1).unwrap();
network.add_connection(&in2, &add1).unwrap();
network.add_connection(&add1, &add2).unwrap();
network.add_connection(&in3, &add2).unwrap();
network.add_connection(&add1, &mul1).unwrap();
network.add_connection(&add2, &mul1).unwrap();
// Register the multiplication as an output
network.as_output(&mul1);
// Insert a closed loop (Positive feedback loop)
network.close_loop(&mul1, &add2);
// Generate the network
network.build();
// Main loop operation
// Update the inputs
network.set_inputs(| index | {
return 5.0;
});
// Iterate over the network
network.process();
let res = network.get_outputs();
// Update the inputs
network.set_inputs(| index | {
return 10.0;
});
// Process twice to let the closed loop propagate
network.process();
let res = network.get_outputs();
It is likely that the library does not implement all the operations that will be needed in your application. Implementing a new operation is simple and can be added outside of the library.
As example: Lets say you want to add a new operation that drops every third element.
// we need a possibility to count. So we need to add a count variable to the struct which will
// be boxed and moved into the node
use sac_base::network::node::Node;
use std::any::Any;
use sac_base::network::network::Network;
pub struct Third {
count: usize,
}
// Now define the functionality
impl Third {
pub fn new<T>() -> Node<T>
where T: Copy + Default
{
// Generate the storage which is the defined struct in a boxed pointer which will be
// moved into the node
let storage = Box::new(Third{
count: 0,
}) as Box<dyn Any>;
Node::new(storage, |data_input: (Vec<&[T]>, usize), data_buffer: &mut Box<dyn Any>, output: &mut Vec<T>| {
// Get the storage back so we can access the count variable
let storage: &mut Third = data_buffer.downcast_mut::<Third>().unwrap();
let (inputs, max_len) = data_input;
// Start iterating over the inputs. Here we only care about one input,
// so we only take 1.
inputs.into_iter().take(1).into_iter().for_each(|data| {
// Iterate over all the values of the input slice.
data.into_iter().take(max_len).into_iter().for_each(|v| {
// Modulo the count
storage.count = storage.count % 3;
// Now we are ready to implement the functionality.
// The content of the vector will be passed into the next nodes as slice.
// We are only working with one sample, so we can just put it into the
// 0 position.
output.clear();
if storage.count == 2 {
// Write default in to the vector if its the third sample.
output.insert(0, T::default());
} else {
// Otherwise just use the normal value
output.insert(0, *v);
}
// Increment the count and modulo it.
storage.count = storage.count + 1;
})
})
})
}
}
let mut network: Network<f32> = Network::new();
// Insert the new node into the network
let third = network.add_operation(Third::new());
network.build();
network.set_input_slice(&third, &[1.0, 2.0, 3.0][..]);
network.process();
let res = network.get_output(&third).unwrap();
assert_eq!(res, [1.0, 2.0, 3.0]);