Accurate channel state information (CSI) is fundamental to reliable and efficient wireless communication. However, traditional pilot-based channel estimation introduces considerable overhead, particularly in massive multiple-input multiple-output (MIMO) systems. This paper investigates pilot-free channel inference to estimate complete CSI directly from multimodal sensing observations, including camera images, LiDAR point clouds, and GPS coordinates. Specifically, we propose a multimodal flow matching framework that fuses heterogeneous sensing modalities into a latent distribution in the channel space, and learns a velocity field that continuously transports samples from the source latent distribution toward the target channel distribution. To shorten and straighten the transport trajectory, we introduce a modality alignment loss that not only regularizes the encoded source distribution but also encourages it to align with the target channel distribution. During inference, the learned flow is efficiently approximated using a second-order numerical integrator. For evaluation, we construct a multimodal simulation dataset with Sionna and Blender, enabling realistic modeling of sensing scenes and wireless propagation. System-level experiments demonstrate that the proposed approach outperforms both pilot-based and sensing-aided baselines in channel estimation accuracy under dynamic environments.

Integrated data and energy transfer (IDET) is considered as a key enabler of 6G, as it can provide both wireless energy transfer (WET) and wireless data transfer (WDT) services towards low power devices. Thanks to the extra degree of freedom provided by fluid antenna (FA), incorporating FA into IDET systems presents a promising approach to enhance energy efficiency performance. This paper investigates a FA assisted IDET system, where the transmitter is equipped with multiple FAs and transmits wireless signals to the data receiver (DR) and the energy receiver (ER), both of which are equipped with a single traditional antenna. The switching delay and energy consumption induced by port selection are taken into account in IDET system for the first time. We aim to obtain the optimal beamforming vector and the port selection strategy at the transmitter, in order to maximize the short-term and long-term WET efficiency, respectively. The instant sub-optimal solution is obtained by alternatively optimizing the beamforming vector and port selection in each transmission frame, while a novel constrained soft actor critic (C-SAC) algorithm is proposed to find the feasible policy of port selection from the long-term perspective. Simulation results demonstrate that our scheme is able to achieve greater gain in terms of both the short-term and long-term WET efficiency compared to other benchmarks, while not degrading WDT performance.

In the mobile network, digital twin (DT)-aided artificial intelligence (AI)-empowered link control is vital to enhance the performance of wireless communication. This paper proposes a deep reinforcement learning (DRL)-convex optimization enhanced time-frequency domain power allocation scheme to reduce the long-term average bit error rate (BER) in multi-user orthogonal frequency division multiplexing (OFDM) systems. To alleviate performance loss caused by trial-and-error during the training period of DRL algorithms, we design a novel practical DT-aided “prediction-then-decision” autonomous wireless link control framework considering the periodic interaction mechanism between the DT and its physical counterpart. A Transformer-based channel generator Mucomformer is implemented in the DT layer to generate large amounts of multi-user virtual channel state information (CSI) in future transmission frames. In addition, the DRL agent is trained over the DT channel in advance and executed in the real-world OFDM system to generate the optimal transmission strategy by considering the interaction mechanism between the DT and the physical counterpart. The simulation results demonstrate that the proposed Mucomformer has lower average prediction error of 2.51 dB compared to the Transformer baseline. The DRL and convex-based power allocation scheme further outperforms the classic strategy. Moreover, the practical DT-aided autonomous link control framework effectively mitigates the performance impairment, achieves an average BER performance gain 45.65% higher than that without DT and achieves faster convergence during the whole training period.

Modulation scheme and power control simultaneously impact the performance of integrated data and energy transfer (IDET). Therefore, some efforts have been invested in deep reinforcement learning (DRL) algorithms to realize adaptive modulation (AM) and adaptive power control (APC), in order to achieve long-term performance improvement. However, the optimal DRL algorithm design for the long-term performance optimization having long-term constraints is still a challenge, while the optimal patterns of IDET-oriented joint AM and APC are not fully understood. This paper aims to maximize the long-term performance of energy harvesting (EH), while satisfying the long-term constraints of spectrum efficiency, bit-error-rate and transmit power, by jointly optimizing the modulation selection and transmit power allocation. Then, a novel DRL algorithm, named constrained parameterized action deep deterministic policy gradient (C-PADDPG), is proposed to find the feasible policy of joint AM and APC for the transformed constraint satisfaction problem. Meanwhile, the optimal policy is searched for via bisection method. Simulation results demonstrate that our solution can achieve significant gain on the long-term EH performance, compared to the traditional genetic algorithm-based solution and other DRL benchmark. Moreover, the communication-efficient and EH-efficient patterns of joint AM and APC generated by the C-PADDPG algorithm are explicitly illustrated and analyzed.