Abstract:In this paper, we present an event-triggered distributed optimization approach including a distributed controller to solve a class of distributed time-varying optimization problems (DTOP). The proposed approach is developed within a distributed neurodynamic (DND) framework that not only optimizes the global objective function in real-time, but also ensures that the states of the agents converge to consensus. This work stands out from existing methods in two key aspects. First, the distributed controller enables the agents to communicate only at designed instants rather than continuously by an event-triggered scheme, which reduces the energy required for agent communication. Second, by incorporating an integral mode technique, the event-triggered distributed controller avoids computing the inverse of the Hessian of each local objective function, thereby reducing computational costs. Finally, an example of battery charging problem is provided to demonstrate the effectiveness of the proposed event-triggered distributed optimization approach.
What problem does this paper attempt to address?
This paper attempts to solve the problems of excessive consumption of communication resources and high computational cost in Distributed Time - Varying Optimization Problems (DTOPs). Specifically, the paper proposes a distributed optimization method based on an event - triggering mechanism to reduce the continuous communication requirements between nodes in a distributed system and avoid calculating the inverse of the Hessian matrix of the local objective function, thereby reducing the computational complexity.
### 1. **Research Background and Problem Description**
As infrastructure such as power systems, transportation networks, and communication networks gradually transform into complex dynamic network systems, time - varying optimization has become a key technology for real - time decision - making. Compared with static optimization, time - varying optimization can better describe the real - time characteristics of certain practical engineering tasks, and thus has been widely used in fields such as multi - rotor UAV transportation, battery charging balance, and multi - robot navigation.
However, existing Distributed Neurodynamic (DND) methods usually require continuous communication between nodes, which leads to excessive consumption of communication resources. In addition, many algorithms need to calculate the inverse of the Hessian matrix of the local objective function, which is very computationally expensive in high - dimensional systems.
### 2. **Main Contributions of the Paper**
To solve the above problems, this paper proposes the following innovations:
1. **Event - Triggering Mechanism**: For the first time, the event - triggering mechanism is introduced into the distributed neurodynamic optimization method, and a communication - efficient distributed controller is designed. Compared with methods with continuous communication, this method significantly reduces the communication frequency and saves communication resources.
2. **Improvement in Computational Efficiency**: By introducing the integral mode technique, the calculation of the inverse of the Hessian matrix of the local objective function is avoided, reducing the computational complexity. Compared with other methods, the controller proposed in this paper only needs to transmit Hessian matrix and gradient information, reducing the amount of calculation.
3. **Simplification of Implementation Conditions**: Compared with estimator - based methods, the method in this paper requires fewer variables and less interaction between nodes, and does not need to assume that the second - order time derivative and the time derivative of the Hessian matrix are bounded, further relaxing the implementation conditions.
### 3. **Theoretical Analysis and Verification**
The paper proves the effectiveness of the proposed method through strict mathematical derivations, including:
- **Consensus and Convergence**: Under certain conditions, the states of all nodes can reach an agreement and track the optimal trajectory.
- **Avoidance of Zeno Behavior**: By designing a reasonable triggering function, it is ensured that there will be no infinitely frequent triggering of events.
### 4. **Numerical Simulation**
To verify the effectiveness of the proposed method, the paper takes the State - of - Charge (SOC) balance problem of the battery energy storage system as an example for simulation. The results show that this method can significantly reduce the consumption of communication and computational resources while ensuring the optimization performance.
### 5. **Conclusion and Outlook**
This paper proposes a distributed neurodynamic optimization method based on an event - triggering mechanism, which effectively solves the problems of communication resource waste and high computational complexity in DTOPs. Future research will further promote this method to make it applicable to a wider range of distributed time - varying optimization problems, especially relaxing the requirement for the consistency of the Hessian matrix of the local objective function.
Through these improvements, this paper provides a more efficient and practical solution for distributed time - varying optimization problems.