Rollup stands out as one of the most effective techniques for blockchain Layer-2 scaling. By processing transactions off-chain, it significantly enhances throughput. However, the predominant Rollup implementations currently rely on centralized sequencers, exposing the system and users to censorship attacks and risking network paralysis. While fully decentralized sequencers encounter latency issues and reduced throughput during the consensus phase. We propose a multi-slot weighted leader election algorithm based on shared sequencers, apply the Proposer-Builder Separation (PBS) model, and use the Fuzzy Cognitive Map (FCM) to analyze and optimize the important influence parameters. With its low trust dependence and high functionality, the probability of selecting malicious nodes is reduced. The sequencing and consensus are separated, so that the transaction can quickly reach soft confirmation. We implement this algorithm in a shared sequencer prototype. The experimental results show that the proposed algorithm parameter settings are in line with expectations, and the probability of electing malicious nodes is significantly reduced. The TPS of the network can cope with the throughput requirements of the Layer-2.