Deadlock is an intrinsic bottleneck in Distributed Real-Time Database Systems (DRTDBS). Deadlock detection and resolution algorithms are important because in DRTDBS, deadlocked transactions are prone to missing deadlines. We propose an Agent Deadlock Detection and Resolution algorithm (ADCombine), a novel framework for distributed deadlock handling using stationary agents, to address the high overhead suffered by current agent-based algorithms. We test a combined deadlock detection and resolution algorithm that enables the Multi Agent System to adjust its execution based on the changing system load, and that selects its victim transactions more judiciously. We demonstrate the advantages of ADCombine over existing algorithms that use agents or traditional edge-chasing through simulation experiments that measure overhead and performance under a widely vary-