In multi-agent safety-critical scenarios, traditional autonomous driving frameworks face significant challenges in balancing safety constraints and task performance. These frameworks struggle to quantify dynamic interaction risks in real-time and depend heavily on manual rules, resulting in low computational efficiency and conservative strategies. To address these limitations, we propose a Dynamic Residual Safe Reinforcement Learning (DRS-RL) framework grounded in a safety-enhanced networked Markov decision process. It’s the first time that the weak-to-strong theory is introduced into multi-agent decision-making, enabling lightweight dynamic calibration of safety boundaries via a weak-to-strong safety correction paradigm. Based on the multi-agent dynamic conflict zone model, our framework accurately captures spatiotemporal coupling risks among heterogeneous traffic participants and surpasses the static constraints of conventional geometric rules. Moreover, a risk-aware prioritized experience replay mechanism mitigates data distribution bias by mapping risk to sampling probability. Experimental results reveal that the proposed method significantly outperforms traditional RL algorithms in safety, efficiency, and comfort. Specifically, it reduces the collision rate by up to 92.17%, while the safety model accounts for merely 27% of the main model’s parameters.

This work proposed surgical forceps as a sensor that integrated step-coated Fiber Bragg Grating (FBG) for the 3-axial force sensing in the percutaneous spinal endoscopic robot. The step-coated FBG achieved a reflection spectrum with double wavelength peaks, providing more wavelength signals for fault-tolerant decoupling. The temperature and force sensitivity were regulated by adjusting the coating size to reduce decoupling error to 3.68% F.S, solving the span temperature (≥ 20 °C) disturbance from the operating room to the operative area. The optical fiber metallization and laser welding package was proposed to achieve stable connections between the quartz optical fiber and metal forceps. The innovative package made the forceps tolerate 180 °C dry-heat sterilization and aqueous erosion, allowing for sterilized reuse (11 times with error lower than 4.98% F.S) and long-term stability (one month with error lower than 4.36% F.S). A Wavelet Fuzzy Entropy (WFE) and Extreme Learning Machine (ELM) based dynamic fault-tolerant decoupling strategy was proposed. The WFE-ELM reduced the sensor error to 4.42% F.S. under the influence of spectrum chirp noise and single-branch FBG breakage, and the fault-tolerant recovery rate within 10% F.S. error was raised to 40.23%. The designed surgical instruments were integrated with a spinal endoscopic surgical robot to conduct the spine model and pig experiments, verifying its force-sensing effectiveness.

Endoscopy holds a pivotal role in the early detection and treatment of diverse diseases, with artificial intelligence (AI)-assisted methods increasingly gaining prominence in disease screening. Among them, the depth estimation from endoscopic sequences is crucial for a spectrum of AI-assisted surgical techniques. However, the development of endoscopic depth estimation algorithms presents a formidable challenge due to the unique environmental intricacies and constraints within the dataset. This paper proposes a self-supervised depth estimation network to comprehensively explore the brightness changes in endoscopic images, and fuse different features at multiple levels to achieve an accurate prediction of endoscopic depth. First, a FlowNet is designed to evaluate the brightness changes of adjacent frames by calculating the multi-scale structural similarity. Second, a feature fusion module is presented to capture multi-scale contextual information. Experiments show that the average accuracy of the algorithm is 97.03% in the Stereo Correspondence and Reconstruction of Endoscopic Data (SCARED dataset). Based on the training parameters of the SCARED dataset, the algorithm achieves superior performance on the other two datasets (EndoSLAM and KVASIR dataset), indicating that the algorithm has good generalization performance.

Autonomous driving is an advanced field that attracts significant attention and engages numerous researchers. However, relying solely on a single autonomous vehicle (AV) is insufficient to meet the demand of future transportation systems. This necessitates the application of connected and autonomous vehicles (CAVs), whose operation relies on multiagent decision-making technology. Currently, research primarily focuses on simple traffic scenarios. However, unsignalized intersections are frequently encountered in rural areas, characterized by high traffic volume, complex interactions, and significant risks. It is crucial to conduct research on the decision-making of CAVs at unsignalized intersections. To address these issues, the lane-changing decision-making of large-scale AVs at unsignalized intersections is studied in this paper. First, an action filter-based deep Q-network method named AF-DQN is proposed, which enables AVs to effectively filter out potentially hazardous lane-changing actions and execute safe actions. Additionally, a multi-objective reward function that considers multiple factors has been designed, including safety, task achievement, and compliance. Moreover, an exploratory training strategy is introduced to train the multi-agent deep reinforcement learning network model. The strategy facilitates agents to learn through exploration in simple scenarios before solving complex driving tasks in more complex scenarios. Finally, experiments are conducted to validate the effectiveness and superiority of the proposed method. Results show that exploratory training accelerates the model’s training speed and improves training effectiveness. Moreover, the AF-DQN method outperforms the baseline method in terms of safety, efficiency, and adherence to traffic rules.