# Daily Note 03/07/2020
###### tags: `Daily Notes` , `O-RAN`
## Name : Christofel Rio Goenawan
## University : Bandung Institute of Technology (ITB)

---

## Schedule:
1. First Topic Meeting and Goals Explanation.
2. Study more detailed how O- RAN utilizes AI and ML.
3. Read explanation about RIC.
## Outcome :
1. Explain how O- RAN utilizes AI and ML.
2. Explain what and works of RIC.
3. Explain what and works of Non- RT RIC.
## Further Plan :
- Understand detailed work and Acumos AI Platform.
- Understand integration of Acumos AI Platform and O- RAN.
- Understand concept, work and implementation of Kubernetes.

---

## Daily Log

### 1. First Topic Meeting and Goals Explanation. <mark>(10.30)</mark>
- The discussion take place in skype. We discuss about study plan mainly and goals of the topic. From the discussion, we can conclude that the first thing we need to study is
1. O-RAN
2. RIC
3. Acumos
4. Kubernetes
We also need to submit daily notes link to Kevin/Derni so that they can check it.

### 2. Study more detailed how O- RAN utilizes AI and ML.<mark>(11.30)</mark>
- Write detailed explanation of how O- RAN utilizes AI and ML.

### 3. Read explanation about RIC. <mark>(14.00)</mark>
- Watch detailed explanation video in reference [3] and read paper in reference [1].
- Read Ferlinda's notes on Non- RT RIC.

---

## Report
### 1. How utilize AI and ML in Network?

The improvement of Machine Learning and Artificial Intelligence have been so tremendous and touch wide scope of area. From the reference [1], there are 3 main domain where we can utilizes AI and ML in network , that is **network design** , **network optimization** and **RAN algorithms**. The simple diagram can be shown below.
<br>

1. **Network Design**
This domain focuses on improving network deployment's parameters such as the number and location of new cells, the associations of cells to baseband (BB) units, the selection of BB units to form an elastic RAN (E-RAN) configuration, etc. This tasks usually relies on planning tools and the domain knowledge of engineers and is performed rather infrequently, such as when new cells are added to an existing network.

2. **Network Optimization**
The network optimization domain focuses on tuning network hyperparameters.The hyperparameters of the algorithm are tuned to produce, for the same measured input, a different output that is more appropriate for the given scenario. Network hyperparameters are optimized to slowly adapt the RAN algorithms to different network scenarios and conditions and bring the performance of a certain area of the network to improved the performance. Examples include hyperparameters for self-organizing networks algorithms and L3 algorithms (mobility, load balancing and so on) for coordination algorithms (such as coordinated multi-point (CoMP), multi-connectivity, carrier aggregation (CA) and supplementary uplink), and for L1/L2 algorithms.

3. **RAN Algorithms**
This domain focuses on optimizing the L3 to L1 control parameters that directly affect the signal transmitted to/from the user. Examples include connectivity decisions and the allocation to users of resources such as modulation and coding scheme, resource blocks, power and beams.

### 2. How utilize AI and ML in RAN?
Optimizing the RAN by tuning the network hyperparameters is safer and easier than redesigning the RAN algorithms with AI-based solutions, as it consists of an outer control loop that does not modify the RAN basic design itself but only tunes its behavior. The simple diagram can be shown below.


The diagram above show how different network hyperparameter values result in different behaviors for the underlying RAN algorithm.

### 3. What is RIC ?
RAN Intelligence Controller ( usually called "RIC" ) is a logical function that enables control and optimization of RAN elements and resources. RIC can be operated in any RAN operation and optimization procedure like radio connection management, mobility management, QoS management, edge services, interference management, etc. RIC can utilize AI?ML models to support the best effective network opertion. There are two types of RIC in Open RAN architecture that are **Non- Real Time RIC (Non- RT RIC)** and **Near- Real Time RIC ( Near- RT RIC)** . Both have different function and works in O- RAN architectures. The function of both RAN can be seen in diagram below.

<br></br>

Non- RT RIC is **RIC that control and optimize RAN elements and resources like AI/ML workflow including model training and updates, and policy-based guidance of applications/features in Near-RT RIC**. And Near- RT RIC is **RIC that that enables near-real-time control and optimization of RAN elements and resources via fine-grained (e.g. UE basis, Cell basis) data collection and actions over E2 interface**.

### 4. How RIC Works ?
Near- RT RIC works by control traffic of information and data form E2 to A1 ( orchestration & automation ) and vice- versa and controled by some **xApps**. xApps is application designed to run on the Near-RT RIC, is also likely to consist of one or more microservices and at the point of on-boarding will identify which data it consumes and which data it provides. The application is independent of the Near-RT RIC and may be provided by any third party.

The works of Non- RT RIC will be explained in ***chapter 6***.

### 5. Why using RIC ?
RIC makes controlling Open RAN traffic systems much easier and affordable by using software controlling system than traditional hardware configuration that is more difficult , time- consuming and expensive.

### 6. What is The Non-RT RIC (Non- Real Time RAN Intelligence Control) ?
> Writer use Ferlinda's notes on Non- RT RIC as study resource in :
> https://hackmd.io/@ferlinda/SkUU8YXCL

As explained before, Non- RT RIC is a logical function that enables non-real-time control and optimization of RAN elements and resources, AI/ML workflow including model training and updates, and policy-based guidance of applications/features in Near-RT RIC. Non-RT RIC focuses on AI/ML implementation. Non-RT RIC can also be a training host. ML training can be performed offline using data collected from the RIC, O-DU and O-RU. Figure below shows the use of the ML components and terminologies:


From Ferlinda's notes , connection of interfaces connected to Non-RT RIC which includes: A1, O1, O2, and Open Fronthaul M-Plane can be shown as diagram below.


Control Loops:

1. **Loop 1** deals with per TTI msec level scheduling and operates at a time scale of the TTI or above.
2. **Loop 2** operates in the Nea- RT RIC within the range of 10-500 msec and above (resource optimization).
3. **Loop 2** operates in the Near- RT RIC within the range of 10-500 msec and above (resource optimization).

From reference [2] there are 3 famous scenario for AI and ML implementation based on location of **training host** and **inference host**.

1. Deployment Scenario 1.1: Non-RT RIC acts as both the ML training and inference host


2. Deployment Scenario 1.2: Non-RT RIC acts as the ML training host and the Near-RT RIC as the ML inference host


3. Deployment Scenario 1.3: Non-RT RIC acts as the ML training host and the O-CU/O-DU as the ML inference host (for FFS)


The table can be shown as below.


---

## Reference
1. https://www.ericsson.com/en/reports-and-papers/ericsson-technology-review/articles/enhancing-ran-performance-with-ai
2. https://static1.squarespace.com/static/5ad774cce74940d7115044b0/t/5e95a0a306c6ab2d1cbca4d3/1586864301196/O-RAN+Use+Cases+and+Deployment+Scenarios+Whitepaper+February+2020.pdf
3. https://www.brighttalk.com/webcast/16515/359818?utm_source=brighttalk-portal&utm_medium=web&utm_content=Parallel%20Wireless&utm_campaign=webcasts-search-results-feed