|
| Citation | Key elements/techniques | Purpose of article reference | Routing | Cluster size | Topological area | Total number of nodes | Deployment type | Packet size | Clustering | Additional techniques | Alive node | Dead nodes | Energy | Lifetime | Packet-based analysis | Delay/latency | Results and findings |
|
| Arjunan and Sujatha [25] | ACO and fuzzy logic | Network division into unequal clusters | Hybrid routing protocol | Variable (from very small to very large) | 200 m × 200 m | 300 | Random | Not mentioned | Yes | Yes | Yes | Yes | Yes | Yes | — | — | Performed first node die and half node die analysis resulting in the following outcomes |
| (i) Improved lifetime |
| (ii) Eliminated hot spot problem |
| (iii) Balanced energy consumption |
|
| Gowda and Jayasree [26] | Hybrid neural network-based energy-efficient routing | Improved PDR and throughput | HNN_GTOA | Variable | 100 m × 100 m | 500 | Random | 512 bytes | Yes | Yes | — | — | Yes | — | Yes | Yes | The proposed approach achieved the following: |
| (i) Highest PDR of 96% |
| (ii) Minimum delay of 3.7 s |
| (iii) High throughput of 1 Mbps |
|
| Guo et al. [27] | Reinforcement learning | RL-based routing | RL-based routing | Variable | 100 m × 100 m | 100 | Random | 512 bits | Yes | — | Yes | — | Yes | Yes | Yes | — | Energy minimization was achieved using a reward function based defined based on the following elements: |
| (i) Residual energy |
| (ii) Link distance |
| (iii) Hop count |
| Further, Q-learning achieved minimal energy consumption but a relatively lower network lifetime |
|
| John and Rodrigues [29] | Taylor series and crow optimization | Energy and throughput | Multiobjective Taylor Crow optimization (MOTCO) algorithm | Variable | | 50, 100 | Random | Not mentioned | Yes | Yes | Yes | — | Yes | — | Yes | — | Despite of decrease in the energy at higher simulation rounds, MOTCO exemplified the following: |
| Throughput at a rate of 0.6514 (highest among other algorithms) |
| Energy left at a rate of 0.1007 (highest among other algorithms) |
|
| Huynh et al. [30] | Delay constraint energy multihop | CH selection strategy | Energy-efficient cluster-based multihop routing | Variable | 100 m × 100 m | 100 | Random | 30 bytes | Yes | — | — | Yes | Yes | — | Yes | Yes | Designed energy cost function resulted in lowest cost route selection |
| Intercluster multihop routing resulted in energy-efficient data transfer with minimum end-to-end delay |
|
| Elhabyan et al. [28] | Nondominated sorting genetic algorithm II and speed-constrained multiobjective particle swarm optimization | Multiobjective minimization | SMPSO for clustering and routing protocol (SMPSO-CR) | Variable | 100 m × 100 m | 100, 200, 300, 400, 500 | Random | Not mentioned | Yes | Yes | — | — | — | — | Yes | — | Simulation analysis showed that SMPSO-CR exhibited higher average throughput for more than 60% of cases |
| SMPSO-CR minimized the number of active CHs per round that reduced the overall energy consumption |
|
| Chang et al. [29] | Machine-learning-based parallel genetic algorithm (MLPGA), PCA, K-means | Multiobjective optimization | Ultra-reliable and low-latency Internet of Things (uRLLIoT)-based efficient routing | Variable | 100 m × 100 m | 50, 100, 150, 200, 250, 300 | Random | Not mentioned | Yes | Yes | Yes | — | Yes | Yes | — | — | Simulation analysis showed that involvement of PCA had significantly improved the network performance using MLPGA in terms of network lifetime, energy consumption, reliability, and complexity |
|
