Endoluminal interventions are crucial for both the diagnosis and treatment of gastrointestinal diseases. However, conventional endoscopic systems are designed for visual inspection and diagnosis, limiting their capacity to assess critical tissue properties such as texture, stiffness, and structural integrity. These unobserved mechanical signatures may contribute to clinically relevant failure modes, including missed lesions, incomplete resection, and adverse events driven by excessive wall loading. Inspired by the tactile sensing mechanisms of rat whiskers, this study introduces a bionic multichannel whisker system, as an additional sensing modality for endoluminal interventions. The system integrates high-fidelity tactile sensing hardware, and advanced signal processing algorithms, enabling precise and reliable measurements of tissue properties, as well as providing real-time radial force feedback. Calibration via affine transformation ensures cross-channel consistency and compensates for manufacturing and installation variances. Validation across multiple experimental settings, including robotic and manual operation, demonstrates performance in texture discrimination, shape reconstruction, and radial force estimation. The presented whisker system is compact to support integration with endoscopic instruments, providing a practical basis for complementing endoluminal diagnostic workflows and for further development of tactile sensing in surgical robotics.

Reinforcement Learning (RL) based control algorithms can learn the control strategies for nonlinear and uncertain environment during interacting with it. Guided by the rewards generated by environment, a RL agent can learn the control strategy directly in a model-free way instead of investigating the dynamic model of the environment. In the paper, we propose the sampled-data RL control strategy to reduce the computational demand. In the sampled-data control strategy, the whole control system is of a hybrid structure, in which the plant is of continuous structure while the controller (RL agent) adopts a discrete structure. Given that the continuous states of the plant will be the input of the agent, the state–action value function is approximated by the fully connected feed-forward neural networks (FCFFNN). Instead of learning the controller at every step during the interaction with the environment, the learning and acting stages are decoupled to learn the control strategy more effectively through experience replay. In the acting stage, the most effective experience obtained during the interaction with the environment will be stored and during the learning stage, the stored experience will be replayed to customized times, which helps enhance the experience replay process.

The effectiveness of proposed approach will be verified by simulation examples.