A trace-based evaluation of adaptive error correction for a wireless localarea network

Wireless transmissions are highly susceptible to noise and interference. As a result, the error characteristics of a wireless link may vary widely dep ending on environmental factors such as location of the communicating syste ms and activity of competing radiation sources, making error control a diff icult task. In this paper we evaluate error control strategies for a wirele ss LAN. Based on low-level packet traces of WaveLAN, we first show that for ward error correction (FEC) is effective in recovering from bit corruptions and that packet length adjustment can reduce packet truncation. However, a s expected, fixed error control policies can perform very poorly, because t hey either introduce too much overhead in "good" environments or are not ag gressive enough in "bad" environments. We address this problem through adap tive error control, i.e., error control policies that adapt the degree of F EC redundancy and the packet size to the environment. The effectiveness of adaptive error control depends on the characteristics of the error environm ent, e.g., the type of errors and the frequency with which the error enviro nment changes. Our evaluation shows that adaptive error control can improve throughput consistently across a wide range of wireless LAN error environm ents. The reason for this effectiveness is that changes in the error enviro nment are often caused by human mobility-related events such as the motion of a cordless phone, which take place over seconds, while adaptation protoc ols can respond in tens of milliseconds. Evaluating adaptive error control in a wireless environments is challenging because repeatable experiments ar e difficult: the wireless environment cannot easily be isolated and the ada ptation process itself changes the environment, which may make trace-based evaluation difficult. We introduce a trace-based evaluation methodology tha t deals appropriately with changes in packet content and size.