Cooperative Intelligent Transport Systems (C-ITS) are pivotal in driving the future mobility towards more autonomous, efficient, and safer roads. These systems leverage communication technologies, such as European Telecommunications Standards Institute (ETSI) ITS-G5 and Cellular Vehicle-to-Everything (C-V2X), to enhance road safety and traffic flow. Central to C-ITS is the concept of Cooperative Perception (CP), enabling Cooperative and Autonomous Vehicles (CAVs) to share data gathered by onboard sensors like cameras, lidar, and radar with nearby CAVs. This cooperation is facilitated through the exchange of Cooperative Perception Messages (CPMs) via V2X communication.

Building upon the concept of CP, Cooperative and Augmented Perception (CAP) significantly enhances the perceptual capabilities of CAVs. A key development in CAP is its integration within infrastructure, equipping traffic entities such as Road-Side Units (RSUs), traffic lights, and road signs with advanced sensors. These sensors not only offer broaden perception capabilities but also share additional perceptual information with CAVs, facilitating an advanced cooperative data fusion process. This process, supported by edge and cloud computing technologies, enables a more comprehensive understanding of the driving environment.

The primary goal is to boost environmental awareness about road objects for CAVs, particularly in complex environments and under various network conditions. However, realizing the full potential of CAP and ensuring its continuous operation face significant challenges. A major challenge lies in managing and filtering the increasing volume of perception data to share via V2X. This data, often containing inaccurate, irrelevant, and redundant information, can overwhelm the communication network and lead to an increased congestion. Addressing these challenges require efficient data filtering strategies that can not only manage the data volume but also to fulfill the specific needs of receiving CAVs for relevant information and dynamically adapt to the varying complexity of the driving environment and changing network congestion levels and control mechanisms.

Motivated by the need to overcome these challenges, this thesis focuses on enhancing the resilience of CAP from networking and communication perspectives, leveraging recent advancements in Deep Reinforcement Learning (DRL) for adaptive behaviors across various complex scenarios with fluctuating conditions.