Source-linked AI summary
Optimal Use of Current and Outdated Channel State Information - Degrees of Freedom of the MISO BC with Mixed CSIT
Tiangao Gou, Syed A. Jafar
TL;DR
The paper asks how imperfect current CSIT and perfect delayed CSIT should be jointly used in a two-user MISO broadcast channel. It adds private messages to a prior mixed-CSIT scheme and derives a tight outer bound, characterizing the optimal DoF while removing the requirement of statistically equivalent user fading.
Problem
The paper addresses characterization of the optimal degrees of freedom for a MISO broadcast channel with imperfect current CSIT and perfect delayed CSIT.
Method
The authors improve the prior mixed-CSIT scheme by incorporating additional private messages and derive an information-theoretic DoF outer bound.
Results
The new achievable DoF is optimal, and the outer bound does not require statistically equivalent fading distributions for the two users.
Takeaways & Limitations
Mixed current and delayed CSIT can be used jointly to identify the DoF-optimal transmission strategy, with a stronger conclusion also in the delayed-CSIT setting.
Takeaways & Limitations
The DoF outer-bound argument relies on i.i.d. temporal channel variations, although the achievable scheme works under the same assumptions as the prior scheme.
Abstract
from arXiv · showhide
We consider a multiple-input-single-output (MISO) broadcast channel with mixed channel state information at the transmitter (CSIT) that consists of imperfect current CSIT and perfect outdated CSIT. Recent work by Kobayashi et al. presented a scheme which exploits both imperfect current CSIT and perfect outdated CSIT and achieves higher degrees of freedom (DoF) than possible with only imperfect current CSIT or only outdated CSIT individually. In this work, we further improve the achievable DoF in this setting by incorporating additional private messages, and provide a tight information theoretic DoF outer bound, thereby identifying the DoF optimal use of mixed CSIT. The new result is stronger even in the original setting of only delayed CSIT, because it allows us to remove the restricting assumption of statistically equivalent fading for all users.
1 Introduction
The paper studies how to optimally combine imperfect current CSIT with perfect delayed CSIT in a two-user MISO broadcast channel. It improves prior mixed-CSIT schemes with additional private messages and establishes a tight DoF characterization that also relaxes user-statistical assumptions in the delayed-CSIT setting.
- 1 Introduction: The model combines imperfect current CSIT with perfect delayed CSIT in a two-user, two-antenna MISO broadcast channel.The channels vary independently across time and users, while users need not have identical fading distributions.
- 1 Introduction: For 0 < α < 1, zero-forcing achieves 2 − 2α DoF and MAT achieves 4/3 DoF, while prior mixed-CSIT work achieves 2(1+α)/(1+2α) DoF.Here α measures the residual signal power after imperfect zero forcing; α = 0 corresponds to perfect current CSIT and α = 1 to no current CSIT.
- 1.1 Overview of results: The largest gain over the better of zero-forcing and MAT is 4/9 at α = 1/3.The achievable DoF curve runs from sum DoF 2 at α = 0 to 4/3 at α = 1.
- 1.1 Overview of results: A tight information-theoretic outer bound proves the new DoF optimal and remains valid without statistically equivalent fading across users.The result is therefore stronger than the earlier delayed-CSIT outer bound, which required statistically equivalent users.
- 1.3 New (optimal) Scheme: Improving upon the achievable scheme of [1]: The new scheme uses partial zero forcing in every transmission phase, adding private messages alongside the quantized common interference.In the second phase, the additional private symbols use zero-forcing directions and power P^−α, arriving at power P^(1−α) for intended receivers and at the noise floor for unintended receivers.
- 1.3 New (optimal) Scheme: Improving upon the achievable scheme of [1]: Over 3 time slots, each user decodes four desired symbols and achieves (3−α)/3 DoF.The scheme combines common-information multicasting with private transmissions across its three phases.
2 System Model
The paper models a two-user, two-antenna MISO broadcast channel with perfect delayed CSI and imperfect current CSI, then transforms it into an equivalent simpler form without changing DoF.
- Channel assumptions: The channel model assumes independent continuous estimated channels and errors, with i.i.d. variation over time and potentially different user distributions.
- CSIT model: The transmitter knows past channel realizations perfectly and current channel estimates imperfectly, while receivers know instantaneous channels perfectly.
- CSIT quality: The parameter α′ measures current-channel estimation quality: α′ = 0 means no current CSI, while α′ ≥ 1 is DoF-equivalent to perfect current CSI.
- Model scope: The paper focuses on optimal use of current and outdated channel information while assuming i.i.d. temporal fading, unlike work that incorporates temporal correlations.
- Equivalent model: Invertible transmitter and receiver transformations align effective transmit dimensions with directions orthogonal to the estimated channels, preserving the channel’s DoF.
3 Results
For the mixed-CSIT MISO broadcast channel, the paper establishes a DoF result together with its optimality. Its outer bound also applies when users have different fading distributions, strengthening the delayed-CSIT result.
- Main result: The main theorem establishes both the achievable DoF and its optimality for the mixed-CSIT MISO broadcast channel.
- Outer bound: The outer bound is the first non-trivial outer bound for the mixed-CSIT setting.
- Outer bound: The outer bound does not require statistically equivalent users, so users may have different fading distributions.
- Delayed-CSIT implication: This removes the statistically equivalent fading requirement from the original delayed-CSIT setting and strengthens that result.
4 Outer bounds
The paper derives outer bounds using physically degraded and compound-channel constructions, obtaining a tight DoF characterization for delayed and mixed CSIT settings. The mixed-CSIT bound varies linearly with current-CSIT quality.
- Delayed CSIT setting: The outer-bound proof uses a compound-channel approach and removes the statistically equivalent fading assumption for delayed CSIT.The argument introduces physically degraded channels and fictional receiver outputs while preserving an outer-bound-consistent capacity relaxation.
- Delayed CSIT setting: Providing one receiver’s output to the other creates a physically degraded memoryless broadcast channel, where delayed-CSIT feedback can be eliminated.The construction relies on the fact that feedback does not increase capacity for physically degraded memoryless broadcast channels.
- Compound-channel construction: The compound setting uses two possible channel realizations per user and requires only linear independence between those realizations.The outer bound therefore applies even when transmitter uncertainty is reduced to a one-bit choice between two independent realizations.
- Mixed CSIT setting: 2nR1 + nR2 ≤ 2n log(P) + n(1 −α) log(P) + n o(log(P)) + o(n) is one asymmetric rate bound in the mixed-CSIT setting.A symmetric construction with User 2 degraded supplies the complementary bound used to obtain the final DoF outer bound.
- Mixed CSIT setting: The final outer bound is a straight line from sum DoF 2 at α = 0 to sum DoF 4/3 at α = 1.This interpolates between the delayed-CSIT and perfect-current-CSIT endpoints in the stated mixed-CSIT model.
5 Achievability
The achievable scheme uses three phases that combine partial zero-forcing, quantized interference multicast, and additional private messages. At high SNR, the latter two phases each require one time slot, enabling four symbols per receiver over three slots.
- Scheme overview: The three-phase achievable scheme uses an equivalent channel model for exposition and combines delayed CSIT with imperfect current CSIT.Its design extends the earlier approach while retaining the same three-phase structure.
- Phase 1: Phase 1 sends four symbols, with a1 and b1 partially zero-forced toward the unintended receivers using imperfect current CSIT.The intended symbols are transmitted at O(P) power, while partial zero-forcing reduces their unintended received power to P^(1−α′).
- Phase 2: Phase 2 delivers quantized common information c1 while simultaneously transmitting one private message for each receiver.The private-message beamformers use partial zero-forcing, and both receivers decode c1 before decoding their private messages.
- Phase 2: At high SNR, phase 2 requires one time slot because t2 → 1 as P →∞.The common information rate is (1 −α′) log P + o(log P), while each phase-2 private message achieves α′ log P + o(log P).
- Phase 3: Phase 3 repeats phase 2’s structure, sending common information c2 and one private message to each receiver.Both private messages achieve α′ DoF, and phase 3 uses t3 = t2 time slots.
- Decoding and DoF: Four symbols are delivered to each receiver over three time slots, after decoding the common information and constructing effective 2 × 2 MIMO channels.The phase-1 symbol pairs achieve rates of (2 − α′) log P + o(log P), with decoding error probability tending to zero.
6 Conclusion
The paper characterizes the optimal DoF of the two-user, two-antenna MISO broadcast channel with delayed and imperfect current CSIT through matching achievability and outer bounds.
- Conclusion: The results identify the DoF-optimal use of outdated and current CSIT in the two-user MISO broadcast channel.The model has two transmit antennas, delayed CSI, and imperfect current CSI.