In this work, the relay aided end-to-end delay optimization in wireless sensor network has been addressed. A routing table creation algorithm was proposed. My major concerns fall into the following aspects.
In the introduction and related work section, the motivation has not been clearly presented. In the related work section, too many words for existing works, on what they have done and their drawbacks. However, the drawbacks that could motivate this paper has not clearly identified. For example, they mentioned that 'However, during the communication process, there is a trade-oﬀ between the data packet size and the transmission power', but there have been many works addresses the same issue that includes trade-off between data packet size and transmission power which leads to optimal power allocation schemes. The new contributions thus are unclear.
In wireless sensor networks, the sensor nodes should be coordinated. In this work, the algorithm seems a distributed implementation. However, if multiple relay nodes obtain the same message, how could de-duplication scheme work? Because by reducing the repeated message transmission, energy can be saved, but link connectivity may suffer. Therefore, there should be trade-off between link connectivity and energy consumption. In this work, the end-to-end delay acts as the main targets, which may be impractical.
The convergence of proposed scheme should be addressed because of the distributed implementation.
For the network or system model, interferences may happen during transmission. For example, are the sensor nodes full duplex? If they are half duplex, interferences can happen, how can you manage this?
There have been numerous studies on the same issue with real-world experiments. Authors are encouraged to construct a testbed to verify the proposed scheme.
What is the controlling overhead in such a network system?
In the simulation, sensor nodes' number is fixed. However, when there are thousands of sensor nodes, can the system still work? When nodes are close to each other, interferences can be very strong. How can you manage this? In this case, the base station should be introduced to coordinate all the sensor noes. Therefore, it is recommended to amend the network model.
This paper proposes a routing protocol to enhance the delay and energy efficiency in WSN for IoT applications. The proposed scheme introduces a novel potential relay information (PRI) metric function which is used to measure the routing potential of nodes. Two-hop information is used to enable nodes to view down their routing path and make better forwarding decisions in real-time to boosts the system performance. The performance is evaluated by simulating.
My suggestions for this work are as follows.
1. Fonts are different in some figures and tables. Please unify the font in this paper. There should be "," after an equation, and "Where" should be "where".
2. What is the limitation of existing methods? It is not presented clearly in Section I.
3. In GreenOrbs, thousands of sensors are deployed in the real environment. There must be a method to deal with the similar problems proposed in this paper. So what is the difference between this paper and the approaches proposed in the other works? What are the advantages of this paper?
4. Why two-hop is studied in this paper, not multi-hop?
5. In the evaluation part, the work  RTLD was published in 2008. Why the author chooses this paper to compare with the method proposed in this paper?
6. More than half of the references in this paper are too old. It is better to reference the state of arts published in recent 5 years. Please reference the following two papers. The format of reference  is not right. There are two "IEEE" in reference .
 Xiulong Liu, Jiuwu Zhang, Shan Jiang, Yanni Yang, Keqiu Li, Jiannong Cao, and Jiangchuan Liu, Accurate Localization of Tagged Objects Using Mobile RFID-Augmented Robots, IEEE Transactions on Mobile Computing (TMC), DOI: 10.1109/TMC.2019.2962129, 2019
 Xiulong Liu, Wenyu Qu, Jia Liu, Alex X. Liu, Jiannong Cao, Jiangchuan Liu, Fast and Accurate Detection of Unknown Tags for RFID Systems – Hash collisions are desirable, IEEE/ACM Transactions on Networking (TON), 28(1): 126-139 (2020)
The manuscript is well written and contains interesting research work.
Attached Commented PDF includes important points to consider.