layout | title | underMenu |
---|---|---|
page |
Plans |
Development |
This page describes what's planned for INET in the future.
-
mobility models
- separate initial positioning from positioning over time
- make them composable sequentially and additively
- extend interface with future movement predictions
-
physical layer
- support capturing a stronger signal while receiving a weaker one
- support aborting an ongoing transmission
- add the simulation of synchronization process (pilot/preamble)
- add missing Weibull and Jakes fading
- add multipath support
- add real CUDA and multi-core parallel execution
-
IPv6
- IPv6 has not yet been through the same refactoring and improvements as the IPv4 model
- after that, xMIPv6 needs to be tested and revised as well
-
NeuralNet based Error model
-
Documentation
- review new and modified NED module and C++ class comments
-
Examples
- review existing examples
- add new more meaningful, more specific feature examples (e.g. crosstalk)
- reduce redundancy in inifiles (e.g. lots of unrelated configuration entries) that makes understanding examples more difficult than necessary
- review directory structure (e.g. more different wireless directories), make the structure more similar to the source directory structure
- extend the wireless tutorial with additional steps
-
Tests
- some examples are rather tests, they were usually created during development for quick testing
- we definitely need more (some?) validation tests (e.g. comparison with other simulators, real-world measurements, or analytical results)
As for the not-so-near future (potentially in a later major version), we are considering some nontrivial changes:
-
synchronous sends
- introducing synchronous (immediate) message sends to reduce communication overhead between protocol layers
-
Factor out the base infrastructure from INET so we would have a subset of components that could be separately supported. Protocol-specific code should be avoided in this codebase.
-
Based on the common infrastructure create a framework where generic network models could be built using only the combination of exsisting (very simple) modules. This could be useful for modeling abstract network concepts.