Interference Detection and Mitigation

Interference detection and mitigation in cooperative multiuser network

In state-of-the-art multiuser multicell communication systems, bandwidth is a limited resource and reused in adjacent cells. A high spectral efficiency is achieved by a network-wide configuration with frequency reuse-one. In these reuse-one networks (RONs), intercell interference (ICI) emerges in the case of a unicast transmission as a critical limiting factor.
An important optimization criterion in network design is a fair distribution of the throughput among the users while maximizing the sum rate of the total network. This is a particular issue in RONs. A network-wide maximization of the signal to interference and noise ratio (SINR), which corresponds to the mitigation of intercell interference, can be achieved by the optimal assignment of the resources in the time-, frequency- and spatial domain.
This project uses beamforming to exploit the spatial domain and temporal user scheduling to exploit the temporal domain. Many previous works on ICI mitigation are based on instantaneous channel state information (CSI). In a coordinated multicell scenario, these strategies become extremely difficult to implement due to the increasing computational complexity and communication overhead for acquiring short-term CSI. The motivation of this project is to optimize the beamforming and the scheduling decisions by exploiting long-term CSI in terms of downlink spatial covariance matrices. Then the optimization of the beamforming weights and scheduling decisions can be done offline by a central unit and reused as long as the long-term CSI is valid. A reduced signaling overhead is the consequence, which is important criterion in a multicell scenario.

We consider two approaches for interference mitigation with long-term beamforming:

  • Beamforming optimization. With max-min beamforming, a fair (balanced) SINR distribution for each scheduling decision can be achieved. This technique is extended to a network-wide coordinated ICI mitigation to achieve a fair distribution of the SINR among the users in the whole network. The beam lobes are jointly optimized to all users inside a cell, which corresponds to a network-wide adaptation of sector patterns. This problem is NP-hard. The beam lobes are constant as long as the long-term CSI is valid; therefore a combination with channel aware scheduling based on locally estimated instantaneous channel state information (CSI) achieves an additional improvement of the throughput.

    Simulation scenario: The green cells have the capability of beamforming. The asteriks denote users distributed in the world.

  • Scheduling optimization. The special case of the sector pattern adaptation problem, where only one user per base station is scheduled is known as the conventional multiuser multicell transmit beamforming problem and can be efficiently solved. An additional advantage of this scheme is that no power is wasted in the direction of the unscheduled users. But then fast beam switching is needed. Using fast beam switching combined with max-min beamforming in each scheduling decision; a balanced SINR is achieved at the expense of a performance degradation of the users with a high SINR. Beside the fairness criterion an improvement of the network-wide sum rate is another optimization criterion. To reduce the tradeoff between fairness and a high network sum rate, network-wide coordinated scheduling decisions avoid that a user is subject to interference by beam lobes of adjacent base stations. These techniques results in a network-wide coordinated joint optimization of beamforming and scheduling, which improve the fairness among the users with a reduced loss of the performance of users with high SINR. A flexible allocation of an arbitrary number of base stations to a user will exploit more degrees of freedom. Especially cell edge users can profit if they are served by adjacent base stations, which further improves the fairness among the users.



Besides the time- and spatial domain, the frequency domain enables additional degrees of freedom for interference mitigation. Different frequency resources can be used to separate users in a way that interference among them can be further mitigated.

All these approaches require accurate knowledge about spatial statistics. The validity of this information depends on the stationary of the channel. Knowledge about stationary intervals will provide information about the duration how long the optimization of the beams is valid and how often the spatial statistics must be estimated and transmitted to the base stations. Finally knowledge about user statistics (throughput and user distribution within the cells) is required for a realistic estimation of the performance gain of the developed algorithms.

The second kind of multiuser networks are cognitive radio (CR) networks. They are inspired by the fact that radio spectrum is a scarce and precious resource in communication systems while it is significantly underutilized with current fixed spectrum assignment policies. CR is proposed as the system which is allowed to access the underutilized spectrum holes in the spectral-temporal-spatial domain, without causing harmful interference to primary users (PUs), which have the priority to occupy such spectrum opportunities. However, CR is not restricted to the concept of spectral-agile radio, but also envisioned to be capable of obtaining and learning the knowledge of its operational and geographical environment, established policies to dynamically and autonomously adjust its operational parameters and protocols. The coexistence of a primary network and cognitive networks is illustrated in Fig. 3. Due to the hierarchical spectrum-sharing structure caused by the coexistence of PUs and CRs, CR inevitably works in the interference-sensitive environments where the interference detection and mitigation become indispensable. We aim at investigating the interference management schemes from the perspective of signal processing in the PHY-layer.


    Apart from the intra-network interference in conventional wireless communication systems, the interference in CR networks contains the inter-network interference from two types of users, i.e., the mutual interference between the primary users and CR networks. The main focus of our research is the detection and mitigation of this inter-network interference. Alternative schemes are exploited depends on the source of the interference.

    • Detection of the interference from PUs to CR networks
      The interference from PU to CR networks can be directly measured by cognitive users through spectrum sensing. The main challenge is that the received primary signal is usually in low SNR region. Furthermore, the detection of primary signal becomes more difficult due to the lack of a priori knowledge such as signal characteristics, synchronization information. We aim at develop efficient sensing algorithms to not only restricted to identify the presence/absence status of primary users, but also to characterize the distribution of primary signals across frequency, time and space domains.
    • Mitigation of the interference from CR networks to PUs
      In the hierarchical spectrum-sharing systems, the maximum amount of interference at the primary receivers should be limited in order to guarantee that the quality of service (QoS) of the primary network is not harmfully degraded by the interfering CR network. The CR transceiver should adopt certain signal processing schemes to mitigate both the cochannel interference and adjacent channel interference cause to primary receivers. We focus on applying precoding and equalization optimization to CR networks for interference management. It could effectively achieve the balance between the interference minimization for PUs and throughput maximization for CR networks. The practical implementation issues should also be considered in this context.   


    Guido Dartmann, Xitao Gong