To address the reliability degradation and accuracy drift of conventional pressure sensors under wide temperature ranges and long-term operation—caused by wire bonding, oil-filled media, and static calibration methods—this paper proposes a high-precision pressure sensor design based on leadless packaging and multi-agent reinforcement learning-based joint compensation. At the signal processing stage, a Multi-Agent Deep Deterministic Policy Gradient (MADDPG) algorithm is introduced to train a multi-channel error compensation model offline using real pressure data, enabling adaptive correction of the sensor output. Experimental results show that the designed sensor achieves basic errors better than 0.1% full scale (FS) across all channels over the full temperature range, and that zero-point and full-scale drift are effectively suppressed under high-temperature long-term operating conditions. The overall performance is significantly superior to conventional least-squares fitting and two-dimensional interpolation calibration methods, demonstrating the effectiveness of the proposed approach in improving the accuracy and long-term stability of pressure sensors.

Multi-Channel Pressure Sensor,Leadless Packaging,Multi-Agent Reinforcement Learning,MADDPG