<p>This study proposes a dynamic lane-control strategy for the novel smart and self-regulating COM-Roundabout, which simultaneously optimizes capacity, safety, and performance under the emergent effects of connected and automated vehicles (CAVs). The proposed model integrates specific capacity and delay models (from the HCM manual) and crash-prediction relationships (from the NCHRP 888 Report) and CAV-specific capacity adjustment factors within a unified optimization platform. Since COM-Roundabout allows the adoption of very different scenarios by activating and deactivating entry and circulating lanes, a dynamic system was considered using one-hour control intervals. The proposed optimization controller balances delays and crash risk while imposing penalties on unnecessary lane switches. Its performance was tested across 5,000 synthetic traffic demand profiles. Compared with a single‑lane baseline, the dynamic scenario reduces mean delays by more than half, while avoiding most of the safety penalties associated with permanent scenarios. Parametric analyses reveal piecewise regime transitions in the delay–safety trade-off that arise from discrete changes in lane-activation states selected by the controller. Higher CAV penetration further magnifies the gains by accelerating congestion relief. Overall, the study demonstrates that dynamic per-arm lane control offers a feasible and transitional strategy for delivering safer and more efficient roundabout operations in an era of smart roads and intersections.</p>