Stochastic approximation based scheduling for real-time applications in wireless networks

We consider Down Link (DL) scenario where a Base Station (BS) supports N real-time flows. Each flow requires a hard delay guarantee, i.e., a packet of flow i arriving at time t must depart before t+D_i, where D_i is the delay requirement of flow i. If a packet is not scheduled before its deadline, then it is dropped. Note that on account of random channel states, packet drop due to deadline violation is inevitable. Fortunately, realtime applications can sustain some loss gracefully. We propose a scheduling algorithm which guarantees that the packet loss due to deadline violation for each flow is below its tolerable limit. Our scheduling algorithm is a randomized scheduling policy where probability of choosing a flow depends on the weight computed for each flow. We provide MATLAB simulation results to study the performance of the proposed schemes and compare them with other scheduling algorithms like Earliest Deadline First (EDF) and Greedy algorithms.