Predictive Resource Allocation in the LTE Uplink for Event Based M2M Applications

For certain event based M2M applications, it is possible to predict when devices will or may need to send data on the LTE uplink. For example, in a wireless sensor network, the fact that one sensor has triggered may increase the probability that other sensors in the vicinity may also trigger in quick succession. The existing reactive LTE uplink access protocol, in which a device with pending data sends a scheduling request to the eNodeB at its next scheduled opportunity, and the eNodeB responds with an uplink grant, can lead to high latencies. This is particularly the case when the system utilizes a high scheduling request period (of up to 80ms) to support a large number of devices in a cell, which is characteristic of M2M deployments. In this paper, we introduce, analyze and simulate a new predictive/proactive resource allocation scheme for the LTE uplink for use with event based M2M applications. In this scheme, when one device in a group sends a scheduling request, the eNodeB identifies neighbor devices in the same group which may benefit from a predictive resource allocation in lieu of waiting for those neighbors to send a scheduling request at their next scheduled opportunity. We demonstrate how the minimum uplink latency can be reduced from 6ms to 5ms and how the mean uplink latency can be reduced by greater than 50% (in certain scenarios) using this method.