Intelligent Sensing Based Resource Allocation Using Q‐Learning With Improved Performance in 5G‐V2X Networks

ABSTRACT In cellular vehicular‐to‐everything (C‐V2X) Mode 4, vehicles schedule resources for themselves autonomously, which is defined by semipersistent scheduling. According to the sensing information obtained in side‐link control information (SCI) through the physical side‐link control channel (PSCCH), vehicles select the resources for cooperative awareness message (CAM) transmission and preserve resources for further transmissions. The half‐duplex problem and allocation of overlapped resources induce packet collisions that degrade overall system performance. In this paper, cluster head vehicle (CHV) assisted balanced resource allocation is proposed entitled CHQ‐RA. CHVs are the most stable vehicles, which are determined by each vehicle's stability index. CHV acts as an agent and incorporates a Q‐learning based paradigm to ensure traffic density based dynamic sensing window allocation. It also disseminates the available resources among cluster member vehicles according to their link stability performance and latency requirements aiming at the best exploitation of the available resources. Comprehensive performance evaluation is done in comparison with the prior relevant works to establish the superiority of the CHQ‐RA under diverse scenarios in terms of the packet reception ratio (PRR), update delay (UD) and collision ratio (CR). PRR, CR, and UD performance are analyzed based on varying traffic density, CAM size, and transmitter‐receiver distance.