Robotics education fosters computational thinking, creativity, and problem-solving, but remains challenging due to technical complexity. Game-based learning (GBL) and gamification offer engagement benefits, yet their comparative impact remains unclear. We present the first PRISMA-aligned systematic review and comparative synthesis of GBL and gamification in robotics education, analyzing 95 studies from 12,485 records across four databases (2014–2025). We coded each study’s approach, learning context, skill level, modality, pedagogy, and outcomes (κ =.918). Three patterns emerged: (1) approach–context–pedagogy coupling (GBL more prevalent in informal settings, while gamification dominated formal classrooms [p <.001] and favored project-based learning [p =.009]); (2) emphasis on introductory programming and modular kits, with limited adoption of advanced software (~17%), advanced hardware (~5%), or immersive technologies (~22%); and (3) short study horizons, relying on self-report. We propose eight research directions and a design space outlining best practices and pitfalls, offering actionable guidance for robotics education.

Preference-based reinforcement learning (PbRL) has emerged as a promising paradigm for teaching robots complex behaviors without reward engineering. However, its effectiveness is often limited by two critical challenges: the reliance on extensive human input and the inherent difficulties in resolving query ambiguity and credit assignment during reward learning. In this paper, we introduce PRIMT, a PbRL framework designed to overcome these challenges by leveraging foundation models (FMs) for multimodal synthetic feedback and trajectory synthesis. Unlike prior approaches that rely on single-modality FM evaluations, PRIMT employs a hierarchical neuro-symbolic fusion strategy, integrating the complementary strengths of vision-language models (VLMs) and large language models (LLMs) in evaluating robot behaviors for more reliable and comprehensive feedback. PRIMT also incorporates foresight trajectory generation to warm-start the trajectory buffer with bootstrapped samples, reducing early-stage query ambiguity, and hindsight trajectory augmentation for counterfactual reasoning with a causal auxiliary loss to improve credit assignment. We evaluate PRIMT on 2 locomotion and 6 manipulation tasks on various benchmarks, demonstrating superior performance over FM-based and scripted baselines. Website at https://primt25.github.io/.

Dynamic wireless channel modeling is essential for future communication system design. However, existing methods struggle to capture the long-term non-stationary behavior of real-world channels driven by high mobility, dense deployment, and complex environments in the sixth generation (6G) scenarios. As system-level design depends on channel statistics rather than exact realizations, statistically consistent channel generation offers a more appropriate modeling strategy. In this work, We propose a deep learning-based hybrid framework that shifts the modeling goal from precise point-wise prediction to generating synthetic channel sequences that statistically replicate long-term non-stationary dynamics. Instead of forecasting exact future states, the model reproduces key statistical features such as the evolution of power delay profiles (PDPs), root mean square (RMS) delay spread and wide-sense stationary (WSS) regions. Experimental results demonstrate that the generated sequences closely match reference data, supporting scalable data generation for digital twins and system-level simulations under realistic dynamic conditions.

In diverse application scenarios of wireless communications, such as massive MIMO channel estimation and sparse multi-user detection, recovering structured sparse signals from linear measurement models under non-ideal conditions remains two key challenges. First, existing sparse prior models are not flexible enough to adapt to the diverse structured sparse properties encountered in various application scenarios, including block sparsity, hierarchical sparsity, and group sparsity. Second, the parameter deviations in the measurement matrix and the potential correlation between its elements seriously impair the robustness of traditional compressed sensing (CS) algorithms. To address these challenges, this paper proposes a three-layer iterative optimization architecture based on joint angle-delay sparsity structure (JADSS). At the same time, the message passing and variational bayesian inference (VBI) methods are embedded in the JADSS iterative architecture, and the JADSS-VBI algorithm is proposed, which significantly improves the signal recovery accuracy under non-ideal conditions. The simulation results show that the JADSS-VBI algorithm exhibits excellent channel estimation performance.

Accurate path loss prediction is crucial to evaluation of wireless coverage, and deep learning based path loss prediction has recently received a lot of attention in complex scenarios. Previous studies based on deep learning methods often overlook impacts of radio propagation. Therefore, this paper proposes a deep learning framework enhanced by empirical propagation models, which guides the network learning process through physical constraints to enhance the prediction model. The model uses a convolutional neural network to extract environmental features and analyzes impacts of different empirical models at various network locations on path loss prediction performance. Experimental results show that the deep learning network using the close-in free space reference distance path loss model achieves fairly high prediction accuracy on a 5.9 GHz urban scenario dataset, with a root mean square error as low as 5.63 dB.