In recent years, sensing networks in dynamic environment have combined the presence of specific event in detection and estimation for monitoring multi-agent networks (MANs). However, the existing machine learning (ML)-based algorithms in multi-agent monitoring has drawbacks such as resource constraints and communication overhead for large-scale networks. To overcome these limitations, a multi-agent deep Q-networks (MA-DQNs) model is proposed for monitoring sensor nodes and central control station by enabling multiple agents to collaborate real-time decision-making for detecting specific environmental factors. Initially, Sensor Scope Environmental Monitoring (SSEM) dataset is used for temperature, humidity, soil moisture, and solar radiation to monitor temporal trends. After that, the preprocessing step includes agent role identification and agent clustering to identify clusters of sensors by using density-based spatial clustering of applications with noise (DBSCAN) in similar environmental conditions. Then, for feature selection, alternating direction method of multipliers (ADMMs) is used to optimize coordination between multiple agents by maintaining local privacy and reduced communication overhead. Finally, the experimental results demonstrate that the proposed MA-DQN model attains with greater accuracy (99.45%) as compared to the existing method self-interested coalitional crowdsensing for multi-agent interactive environment monitoring (SCC-MIE).