|
1 | 1 | # What's New in Simu5G |
2 | 2 |
|
| 3 | +## v1.4.3 (2026-02-18) |
| 4 | + |
| 5 | +This release represents a major milestone in the complete overhaul of the Simu5G |
| 6 | +codebase to make it architecturally more compliant with the 3GPP |
| 7 | +specifications, modernize the code, and adopt the best practices of the INET |
| 8 | +Framework on which it is based. The goal is to pave the way for a clean |
| 9 | +implementation of new protocol features such as TSN support. |
| 10 | + |
| 11 | +Key achievements in this release: |
| 12 | + |
| 13 | +- **More explicit Control Plane modeling**: Simu5G is advertised as a User Plane |
| 14 | + simulator, but since it was also used to model dynamic scenarios such as |
| 15 | + handovers, it always contained elements of the Control Plane distributed across |
| 16 | + various modules. The new direction is to make these elements more explicit and |
| 17 | + centralized, such as creating dedicated RRC (Radio Resource Control) and |
| 18 | + Session Management Function (SMF) implementations. It is an explicit non-goal |
| 19 | + to simulate Control Plane messaging -- its functionality will be implemented |
| 20 | + with C++ method calls across modules. Thus, Simu5G remains a User Plane |
| 21 | + simulator, but with the possibility to more faithfully model dynamic |
| 22 | + scenarios with heavy Control Plane involvement. While this goal is not fully |
| 23 | + realized in this release, many changes point into that direction. |
| 24 | + |
| 25 | +- **Control info refactoring**: Cleaned up `UserControlInfo` and |
| 26 | + `FlowControlInfo` by removing 5+ unused fields and splitting out smaller, |
| 27 | + focused tags. For example, IPv4 addresses, only used between `Ip2Nic` and PDCP, |
| 28 | + have been factored out into an `IpFlowInd` tag. This improves modularity, |
| 29 | + reduces coupling between protocol layers, and makes the code easier to |
| 30 | + maintain and extend. |
| 31 | + |
| 32 | +- **Added vital missing fields to PDCP and MAC headers**: Protocol layers now |
| 33 | + use proper header fields instead of "tunnelling" information via |
| 34 | + `UserControlInfo` and `FlowControlInfo` packet tags that would not exist in a |
| 35 | + real implementation. For example, PDCP sequence numbers are now carried in |
| 36 | + PDCP headers, and LCIDs are stored in MAC PDU subheaders. This makes the |
| 37 | + simulation more realistic and packet contents more inspectable in Qtenv. |
| 38 | + |
| 39 | +- **Explicit setup of logical connections instead of on-the-fly discovery**: |
| 40 | + This is a key architectural change, which also largely motivated the |
| 41 | + previous items. In previous iterations of Simu5G, data structures associated |
| 42 | + with logical connections / bearers were created in each protocol layer as they |
| 43 | + encountered packets that belonged to new connections. Moreover, part of the |
| 44 | + connection state was carried along by the packets in `FlowControlInfo` tags |
| 45 | + instead of stored inside the protocol. While this modeling approach still |
| 46 | + allowed for faithful simulation of the traffic while keeping the |
| 47 | + implementation simple, it has become a roadblock for implementing complex |
| 48 | + dynamic scenarios where connections come and go. In this iteration, |
| 49 | + centralized session and bearer management (SMF-like functionality) was added |
| 50 | + to the `Binder` module, and RRC modules were added to NICs to carry out local |
| 51 | + configuration. This brings the architecture closer to the 3GPP control/user |
| 52 | + plane separation, making it easier to implement features like handovers |
| 53 | + correctly. This is work in progress: SMF is still part of `Binder` and not a |
| 54 | + separate module, and connection setup is still triggered by the first packet |
| 55 | + of the connection hitting PDCP on the way out. However, moving the SMF code |
| 56 | + into its own module will be trivial, and the single `Binder` method call in |
| 57 | + PDCP can now be easily replaced with static configuration or with calls from a |
| 58 | + more detailed Control Plane implementation. |
| 59 | + |
| 60 | +- **Removed incomplete MIMO support**: Removed MIMO-related code and parameters. |
| 61 | + The existing MIMO code was incomplete (e.g., PMI values were computed but |
| 62 | + never used). Removing it simplifies the codebase and model parameterization, |
| 63 | + and avoids confusion about capabilities. MIMO support will be added in a future |
| 64 | + release, with a different approach. |
| 65 | + |
| 66 | +- **Initialization cleanup**: Reorganized module initialization into well-defined, |
| 67 | + Simu5G-specific init stages. This eliminates hidden cross-module dependencies, |
| 68 | + makes the initialization order explicit and verifiable, and prevents subtle bugs |
| 69 | + caused by modules accessing uninitialized data in other modules. |
| 70 | + |
| 71 | +Further notable changes: |
| 72 | + |
| 73 | +- In UE models, `masterId` and `nrMasterId` were renamed to `servingNodeId` and |
| 74 | + `nrServingNodeId`. The old names were confusing because "Master" has a |
| 75 | + specific meaning in Dual Connectivity (Master eNB vs Secondary gNB), unrelated |
| 76 | + to the UE's serving node. |
| 77 | + |
| 78 | +- In UE models, the `macCellId`, `nrMacCellId` parameters were removed. In |
| 79 | + practice, the code already used the serving node ID as cell ID. |
| 80 | + |
| 81 | +- `macNodeId` assignment was moved to NED, and now it is based on the new |
| 82 | + `simu5g_seq()` NED function that generates an integer sequence. This replaces |
| 83 | + the earlier approach where node IDs were assigned by `Ip2Nic` during |
| 84 | + initialization, and stored back into the module parameters for other modules to |
| 85 | + use. |
| 86 | + |
| 87 | +- `LteRlcPduNewData` and `LteRlcSdu` packet chunks were converted to packet |
| 88 | + tags, as they represent internal metadata rather than actual protocol data. |
| 89 | + |
| 90 | +- In the C++ code, merged the `ENODEB` and `GNODEB` node type enum values into a |
| 91 | + single `NODEB` value, with a separate `isNr` flag where needed. This change |
| 92 | + simplified a large number of "if" conditions throughout the codebase. |
| 93 | + |
| 94 | +To port your existing Simu5G simulations to this version, apply the following |
| 95 | +changes to the ini files: |
| 96 | + |
| 97 | +- Change `masterId` to `servingNodeId` (and `nrMasterId` to `nrServingNodeId`), |
| 98 | + unless it refers to the Master/Secondary distinction in a Dual Connectivity |
| 99 | + setup. |
| 100 | + |
| 101 | +- Remove `macCellId` and `nrMacCellId` parameter assignments for UE modules. |
| 102 | + |
| 103 | +- Delete ini entries that set the following removed MIMO-related parameters: |
| 104 | + `numRus`, `ruRange`, `ruStartingAngle`, `ruTxPower`, `antennaCws`, |
| 105 | + `muMimo`, `pmiWeight`, `lambdaMinTh`, `lambdaMaxTh`, `lambdaRatioTh`, |
| 106 | + `feedbackGeneratorType`. |
| 107 | + |
| 108 | +- For `initialTxMode`, the following values are no longer valid: |
| 109 | + `SINGLE_ANTENNA_PORT5`, `OL_SPATIAL_MULTIPLEXING`, `CL_SPATIAL_MULTIPLEXING`, |
| 110 | + `MULTI_USER`. Remove the parameter assigment to use the default. |
| 111 | + |
| 112 | +There are many more changes that potentially affect existing simulations, and |
| 113 | +projects extending, or built on top of, Simu5G. They cannot all be covered here |
| 114 | +in detail - see the git history for details. |
| 115 | + |
| 116 | + |
3 | 117 | ## v1.4.2 (2025-11-27) |
4 | 118 |
|
5 | 119 | This is primarily a bugfix release. |
|
0 commit comments