Issues and Main Results

Existing wireless scheduling solutions typically take an adaptive approach, in which the scheduler adaptively adjusts its scheduling decisions based on the estimated current channel state. In a mobile network where wireless channel may exhibit a wide range of error patterns, if the estimated channel state is erroneous, the performance gain is uncertain and may degrade significantly. This paper explores an alternative design approach, i.e., robust packet scheduling in wireless cellular networks. It introduces a simple game-theoretic modeling framework in which we can more concretely pose the question of whether the scheduling solution can meet user needs. We take a game theoretic approach in which the scheduler plays a zero-sum game against the channel error adversary. This way, it enables us to derive the worst-case optimal scheduling policies. We have also studied a class of probabilistic scheduling policies. Our study leads some new insights. The analysis shows that under heavy channel error case, the optimal solution to maximizing network total utility does not possess short-fairness property in the presence of heavy channel errors. However, the long-term unfairness may still be bounded. We also find that, probabilistic scheduling may lead to higher worst-case performance compared with deterministic policies. This opens door for future design of probabilistic approach to attack channel errors. Finally, robust scheduling does not mean always conservative and take low-risk options. In fact, the scheduler sometimes would take high risks in hope for high utility gain. We compare our results with adaptive scheduling policies in simulations.

 Project Members

• Student: Xiaoqiao Meng, Zhenghua Fu
• Faculty: Songwu Lu

Publications

• Xiaoqiao Meng, Zhenghua Fu, Songwu Lu, "Robust packet scheduling," to appear in ACM MONET, 2002