Vehicular perception is vital for autonomous vehicles, as it requires precise evaluation of both the external environment and the vehicle’s movements to enhance road safety and driving efficiency. In this work, we focus on developing a comprehensive framework that integrates inertial sensor data with AI models to facilitate vehicular perception through two fundamental components: exteroception and proprioception. The exteroception component is designed to detect and analyze road characteristics, addressing both temporary anomalies such as potholes, cracks, and speed bumps, as well as consistent features such as surface type and overall road conditions. In contrast, the proprioception component is dedicated to understanding vehicle dynamics to recognize driving behaviors, including lane changes, braking, skidding, and turning, while also establishing persistent behavior patterns indicative of safe or hazardous driving using data from accelerometers and gyroscopes. Experiments highlighted the effectiveness of the proposed system, which achieved an accuracy of 93.4% in detecting speed bumps and 92.4% in classifying road surface conditions, thus exceeding existing benchmarks.