Research on Multi-sensor Fusion Localization of Wheel-Legged Robots in Dynamic Scenarios

With the rapid development of human beings in motion control and localization perception, intelligent mobile robots are widely used in various scenarios, such as industrial production, medical care, logistics and distribution. Among them, wheel-legged robots have both the high mobility of wheeled robots and the high adaptability of legged robots, which can not only move fast on platform terrain, but also adapt to rugged and complex ground with the multiple degrees of freedom of their legs. Wheel-legged robots face the following main problems in high-precision robust localization in multiple scenarios: there are a lot of dynamic factors in daily indoor and outdoor scenarios, such as vehicles, pedestrians, etc., which require specific processing of dynamic environment in the visual front-end part to improve the effect of visual factor constraints; as ground mobile robots, the localization framework of visual inertial fusion (VIO) will show scale degeneration phenomenon, resulting in long-distance localization accuracy decline. In addition, as a two wheeled inverted pendulum robot, the ground constraint form will undergo corresponding changes. Aiming at the above problems, this paper takes the hydraulic-driven wheel-legged robot as the object and proposes a multi-sensor fusion localization method in dynamic scenarios.