Recent achievements in the field of computer vision, reinforcement learning, and locomotion control have largely extended legged robots’ maneuverability in complex natural environments. However, little research focuses on sensing and analyzing the physical properties of the ground, which is crucial to robots’ locomotion during their interaction with highly irregular profiles, deformable terrains, and slippery surfaces. A biomimetic, flexible, multimodal sole sensor (FMSS) designed for legged robots to identify the ontological status and ground information, such as reaction force mapping, contact situation, terrain, and texture information, to achieve agile maneuvers was innovatively presented in this paper. The FMSS is flexible and large-loaded (20 Pa–800 kPa), designed by integrating a triboelectric sensing coat, embedded piezoelectric sensor, and piezoresistive sensor array. To evaluate the effectiveness and adaptability in different environments, the multimodal sensor was mounted on one of the quadruped robot’s feet and one of the human feet then traversed through different environments in real-world tests. The experiment’s results demonstrated that the FMSS could recognize terrain, texture, hardness, and contact conditions during locomotion effectively and retrain its sensitivity (0.66 kPa−1), robustness, and compliance. The presented work indicates the FMSS’s potential to extend the feasibility and dexterity of tactile perception for state estimation and complex scenario detection.

Some problems associated with building map in confined environments for mobile robots are studied in this work. The uncertainties arising from specular reflection of ultrasonic sensors cannot make mobile robots recognize their surrounding environments correctly when mobile robots navigate in a confined environment. If the environment is known and unchanged, its map can be built with ultrasonic sensors in advance before navigation is performed. This map can provide reliable information about the environment when mobile robots work. A multi-layer fusion algorithm is proposed for building the map in advance. For the situation that the environment is unknown or variable, an adaptive ultrasonic sensor model is applied to build map dynamically. The experiment results indicate that the above fusion algorithms improve the performance of ultrasonic sensors.

No abstract

No abstract