Offline Zero-Shot Reinforcement Learning requires an agent to solve unseen tasks using only a fixed offline dataset without explicit rewards. A central challenge is learning representations that capture both high-level long-term planning and low-level physical dynamics. We propose a novel framework, Dynamics-Aware Planning Representation (DAPR), which disentangles these two aspects via complementary contrastive objectives. Specifically, DAPR learns goal-oriented planning directions and local dynamics-consistent directions in the latent space. By jointly enforcing these constraints, DAPR yields representations that balance “where to go” with “how to move.” Experiments on standard locomotion benchmarks (Walker, Cheetah, Quadruped) demonstrate that DAPR consistently improves performance and generalization over strong baselines, achieving substantial gains on precision demanding tasks.

Dynamic head pruning in Vision Transformers (ViTs) improves efficiency by removing redundant attention heads, but existing pruning policies are often difficult to interpret and control. In this work, we propose a novel framework by integrating Sparse Autoencoders (SAEs) with dynamic pruning, leveraging their ability to disentangle dense embeddings into interpretable and controllable sparse latents. Specifically, we train an SAE on the final-layer residual embedding of the ViT and amplify the sparse latents with different strategies to alter pruning decisions. Among them, per-class steering reveals compact, class-specific head subsets that preserve accuracy. For example, bowl improves accuracy (76%→82%) while reducing head usage (0.72→0.33) via heads h2 and h5. These results show that sparse latent features enable class-specific control of dynamic pruning, effectively bridging pruning efficiency and mechanistic interpretability in ViTs.

Inertial measurement unit (IMU) is a critical component for autonomous robot navigation. This paper presents a novel training framework that uses knowledge distillation to enhance low end, or low precision, IMU performance by transferring learned representations from high end IMU. In the proposed framework consisting of dual branch, teacher branch trained with high end IMU sensor data is used to train the student branch with low end IMU. For training the low end IMU branch, combined loss function that balances signal matching and knowledge transfer is used. Testing with the MAGF-ID dataset demonstrates substantial reduction of mean absolute error as much as 39-65% with accelerometer measurements and 47-70% for gyroscope readings with best improvement in the (gyroscope) X-axis, as compared to original low end IMU measurements. Qualitative analysis also reveals that proposed framework is efficient in systematic error correction and noise suppression.

Vehicular accidents pose a substantial risk to drivers, underscoring the persistent and vital need for heightening safety measures. Early accident anticipation mechanisms are imperative for proactive measures, while detection accuracy is pivotal for prompt response and effective post-accident mitigation. Accurate and early anticipation of accidents for automated driving assistance systems in vehicles or CCTV in cities remains a complex task due to the intricate spatial-temporal interactions within traffic videos. This study presents text-informed magnitude enhancement in contrastive multiple-instance feature learning for vehicle accident detection and anticipation (TIME-VAD). Text is a better representative of concepts when compared to images in video, thus multi-modal learning is suitable. Also, the traditional assumption about feature magnitude of accidents and normal frames in magnitude based multiple-instance learning using weak supervision may not hold. This has led to the development of a novel weak-supervised learning strategy involving magnitude enhancement from textual concepts. For a better frame-level perception of accident risks in videos, dynamic temporal attentions are refined using the proposed dilated temporal conv-attention (DTCA) block. In-depth component-level analysis is performed to showcase the model’s efficacy while elucidating its operational mechanisms. Evaluation is conducted on three benchmark datasets, considering both earliness and accuracy-related metrics. Extensive experiments demonstrate that our TIME-VAD model outperforms the existing models. Compared to the previous top-performing supervised model achieving 84.7% accuracy, TIME-VAD achieves a 94.44% accuracy (measured by ROC-AUC) on the largest DoTA dataset. Notably, our model also excels in measuring how early it detects accidents compared to previous methods. The code will be released onhttps://github.com/sumitmishra209/TIME-VAD

In multi-goal reinforcement learning with a sparse binary reward, training agents is particularly challenging, due to a lack of successful experiences. To solve this problem, hindsight experience replay (HER) generates successful experiences even from unsuccessful ones. However, generating successful experiences from uniformly sampled ones is not an efficient process. In this paper, the impact of exploiting the property of achieved goals in generating successful experiences is investigated and a novel cluster-based sampling strategy is proposed. The proposed sampling strategy groups episodes with different achieved goals by using a cluster model and samples experiences in the manner of HER to create the training batch. The proposed method is validated by experiments with three robotic control tasks of the OpenAI Gym. The results of experiments demonstrate that the proposed method is substantially sample efficient and achieves better performance than baseline approaches.

Optical character recognition(OCR) is the technology to identify text characters embedded within images. Conventional OCR models exhibit performance degradation when performing with noisy images. To solve this problem, we propose a novel model, which combines computer vision using optical sensor with natural language processing by bidirectional encoder representations from transformers(BERT) and cosine similarity scoring. The proposed model uses a confidence rate to determine whether to utilize optical sensor alone or BERT/cosine similarity scoring combined with the optical sensor. Experimental results show that the proposed model outperforms approximately 4.34 times better than the conventional OCR.

Efficient exploration for an agent is challenging in reinforcement learning (RL). In this paper, a novel actor-critic framework namely virtual action actor-critic (VAAC), is proposed to address the challenge of efficient exploration in RL. This work is inspired by humans' ability to imagine the potential outcomes of their actions without actually taking them. In order to emulate this ability, VAAC introduces a new actor called virtual actor (VA), alongside the conventional actor-critic framework. Unlike the conventional actor, the VA takes the virtual action to anticipate the next state without interacting with the environment. With the virtual policy following a Gaussian distribution, the VA is trained to maximize the anticipated novelty of the subsequent state resulting from a virtual action. If any next state resulting from available actions does not exhibit high anticipated novelty, training the VA leads to an increase in the virtual policy entropy. Hence, high virtual policy entropy represents that there is no room for exploration. The proposed VAAC aims to maximize a modified Q function, which combines cumulative rewards and the negative sum of virtual policy entropy. Experimental results show that the VAAC improves the exploration performance compared to existing algorithms.

Robust monocular depth estimation is essential for various applications relying on visual cues to understand the real world. To ensure robustness, unsupervised domain adaptation is widely used for monocular depth estimation. Despite recent advances, existing methods often struggle in outdoor environments due to adverse environmental conditions and limited datasets. Intentionally corrupted images obtained from real images captured in clear weather conditions for unsupervised domain adaptation often fail to accurately represent the complex characteristics of diverse environments, leading to unrealistic training data. From this viewpoint, simulation data offering more plausible representation of adverse weather conditions are used. However, it still presents drawbacks due to potentially degrading adaptation capabilities. To address the limitations of using simulation data, we propose a wild-condition pass filtering module that extracts wild-condition features and captures cross-domain relationships from both real and simulation datasets. This enables comprehensive learning of different conditions from each dataset and improved performance on real adversarial target images. The proposed method achieves a notable 22% improvement over the baseline on the Foggy Cityscapes dataset, highlighting the importance of employing realistic domain adaptation techniques to effectively address the challenges posed by adverse environmental conditions. The code is available at https://github.com/JH2-LEE/wide.

In urban cities, visual information on and along roadways is likely to distract drivers and lead to missing traffic signs and other accident-prone (AP) features. To avoid accidents due to missing these visual cues, this paper proposes a visual notification of AP-features to drivers based on real-time images obtained via dashcam. For this purpose, Google Street View images around accident hotspots (areas of dense accident occurrence) identified by a real-accident dataset are used to train a novel attention module to classify a given urban scene into an accident hotspot or a non-hotspot (area of sparse accident occurrence). The proposed module leverages channel, point, and spatial-wise attention learning on top of different CNN backbones. This leads to better classification results and more certain AP-features with better contextual knowledge when compared with CNN backbones alone. Our proposed module achieves up to 92% classification accuracy. The capability of detecting AP-features by the proposed model were analyzed by a comparative study of three different class activation map (CAM) methods, which are used to inspect specific AP-features causing the classification decision. The outputs of the CAM methods were processed by an image processing pipeline to extract only the AP-features that are explainable to drivers and notified using a visual notification system. Range of experiments was performed to prove the efficacy and AP-features of the system. Ablation of the AP-features taking 9.61%, on average, of the total area in each image sample increased the chance of a given area to be classified as a non-hotspot by up to 21.8%.

Training agents via deep reinforcement learning with sparse rewards for robotic control tasks in vast state space are a big challenge, due to the rareness of successful experience. To solve this problem, recent breakthrough methods, the hindsight experience replay (HER) and aggressive rewards to counter bias in HER (ARCHER), use unsuccessful experiences and consider them as successful experiences achieving different goals, for example, hindsight experiences. According to these methods, hindsight experience is used at a fixed sampling rate during training. However, this usage of hindsight experience introduces bias, due to a distinct optimal policy, and does not allow the hindsight experience to take variable importance at different stages of training. In this article, we investigate the impact of a variable sampling rate, representing the variable rate of hindsight experience, on training performance and propose a sampling rate decay strategy that decreases the number of hindsight experiences as training proceeds. The proposed method is validated with three robotic control tasks included in the OpenAI Gym suite. The experimental results demonstrate that the proposed method achieves improved training performance and increased convergence speed over the HER and ARCHER with two of the three tasks and comparable training performance and convergence speed with the other one.

Games have been used as excellent testbeds for research on artificial intelligence (AI) and computational intelligence for their diversity and complexity. In this article, we present AI World Cup, a set of AI competitions based on the game of soccer. We provide an introduction to the three challenges that concern a robot soccer match using both value-based and image-based state representations. AI Soccer runs the robot soccer match by participants managing each team of five two-wheeled robots. AI Commentator and AI Reporter observe the AI Soccer match and output real-time commentary and a summary article, respectively. Also, we introduce the AI World Cup platform along with rationale behind notable design choices. The official international AI World Cups held in 2018 and 2019 and the AI Masters competition held in 2019 as a part of the World Cyber Games are briefly discussed. Technical aspects of the strategies developed by participants are also discussed.

Wireless power transfer (WPT) is being developed to supply electric power to electric trains, using a source coil in/on the railway track and a pick-up coil on the train. A number of benefits can be obtained by eliminating the catenary and pantograph currently used for railway electric power supply. However, the WPT employs coils, and the electromagnetic field generated by the WPT can interfere with conventional train detection methods, such as track circuits and RFID, installed on the tracks. Train position information is critical for railway operation, especially for high speed trains. This paper proposes a novel system which provides train position information using a source coil segment. It consists of onboard sensor coils, ferrite blocks designed using specific position information, and a detector. The proposed system can be implemented using the source coil and load coil of the WPT. Information about the train's relative position is obtained by the detection of ferrite blocks distributed within the source coil segment. The train position information provided by the ferrite components is detected by onboard sensor coils and a detector. The proposed ferrite position identification (FPID) system provides accurate train position information and is not affected by WPT electromagnetic interference. The operating principles of the FPID system are presented in detail. The performance of the proposed FPID system was measured by simulations and experiments.

In this paper we propose a novel importance sampling (IS) scheme to estimate the low error-rate performance of low-density parity-check (LDPC) codes over Rayleigh fading channels. The proposed scheme exploits the structural weakness of LDPC codes due to trapping sets (TSs). The Rayleigh fading distribution on the bits belonging to a TS is biased by parameter scaling (PS), while the noise distribution on them is biased via mean translation (MT) according to their fading coefficients. The biases in PS and MT are determined so that the variance of the proposed IS estimator is minimized. The proposed IS scheme is compared with the Monte Carlo (MC) simulator and other IS schemes modified from the conventional IS scheme employed for performance estimation of LDPC codes over an AWGN channel. Numerical results show that it provides much more accurate performance than other IS schemes. Furthermore, the proposed IS estimator is even more efficient than the MC estimator and other IS estimators from the viewpoint of the number of required simulation runs.

Selected mapping (SLM) based schemes effectively reduce the peak-to-average power ratio (PAPR) of orthogonal frequency division multiplexing (OFDM) systems. However, they require side information (SI) transmission, which incurs a loss in the data throughput in addition to the increased system complexity. This paper presents a blind SLM scheme based on a decision metric obtained from pilot sub-channel responses. A novel SI detection method enabling low complexity data decoding is proposed. The SI is detected by exploiting the high autocorrelation between adjacent pilot sub-channel responses. The SI detection error rate is analytically derived and compared with that obtained by simulations. Simulation results of the data decoding scheme based on the proposed SI detection method show the bit error rate performance comparable to that of the simplified maximum likelihood (ML) data decoding scheme, while the computational complexity is close to that of the embedded SI based decoding scheme.

A peak-to-average power ratio (PAPR) reduction based on adaptive all-pass filters (AAPFs) for multiple-input single-output orthogonal frequency division multiplexing systems with space-frequency block coding is proposed. The AAPF creates phase rotation to reduce the PAPR. With the scattered pilot pattern, the AAPF enables data recovery without side information (SI). The PAPR reduction performance and the BER performance with considered system parameters are comparable to those of the blind SLM (B-SLM) scheme and the low complexity SLM (LC-SLM) scheme with SI. The computational complexity is much lower than that of the B-SLM scheme and comparable to that of the LC-SLM scheme.

In this paper, we propose a high performance asynchronous on-chip bus with multiple issue and in-order/out-of-order completion for a Globally Asynchronous Locally Synchronous (GALS) design. The proposed bus implementation can be characterized with distributed and modularized control units based on a layered architecture to support multiple issue and in-order/out-of-order completion. Simulation results reveal that throughputs of asynchronous on-chip buses with multiple issue and in-order/out-of-order completion increases by 31.3% / 34.3%, while power consumption overhead is only 6.76% / 3.98% respectively, compared to a simple asynchronous on-chip bus with only a single issue feature.

Linear microcells have been proposed for PCS systems employing low base station antennas. The layout of linear cells in such a system will depend on the characteristics of the propagation on line-of-sight (LOS) radio links, which have been measured by various groups. For planning purposes, it is common to use a power law dependence of the path loss on distance, together with slow fading statistics, both of which are obtained from the measurements by the regression analysis. The simplest regression analysis fits a single straight line to measurements of path loss (dB) plotted versus the logarithm of the antenna separation. However, it has been found that a two-segment regression fit gives a more accurate representation for the trend of the path loss variation, and as a consequence results in a smaller standard deviation for the slow fading statistics. In this paper, we show how the use of the two-segment regression leads to a more efficient system design employing less base stations to achieve the same quality of service (QOS).