Research On Environmental Perception And Behavior Decision Of Quadruped Robots

In the field of mobile robotics,quadruped robots have multiple advantages compared to other types of robots such as wheeled or tracked robots.They possess high overall capabilities and demonstrate strong adaptability in complex environments,making them highly promising and valuable for research and development.Enhancing the robust motion capabilities and terrain adaptability of quadruped robots has been a continuous goal in this field.Additionally,with the advancement of artificial intelligence and big data,exploring more intelligent quadruped robots to enhance their autonomous perception and decision-making abilities is also an important direction of development.For the entire robot system,the environmental perception layer serves as a crucial hub for the system’s interaction with the external world,playing a pivotal role.It not only serves as a priori information for the control and decision-making layer,by measuring and making a pre-estimation of the surrounding environment,but also serves as the foundation for higher-level tasks such as autonomous navigation and human-robot interaction.The absence of a perception system often limits the overall performance of the robot.Therefore,in order to enhance the quadruped robot’s motion capabilities and its cognitive decision-making abilities,this paper focuses on the following research work regarding the environmental perception system for quadruped robots:To avoid the instability issues in the control system caused by excessive foot-end impulses in traditional control strategies,this paper discusses a local map construction method based on quadruped robots.By integrating real-time data from camera sensors with robot pose data,more accurate environmental information is obtained.Based on this information,foothold decision-making and end-effector trajectory planning are performed,thereby enhancing the robustness of quadruped robot motion.Additionally,in order to better align with real-world scenarios,this paper employs semantic network segmentation to obtain terrain classification and Bayesian recursive estimation to acquire terrain attribute information.These serve as constraints for solving the nonlinear model predictive control,making the control model more consistent with real-world conditions.The robot’s environmental perception system provides prior information during the interaction with the control system,transforming control behavior from a trial-and-error process.In order to enhance the intelligent perception of quadruped robots in their environment,this paper presents a research on vision-based perception optimization and object detection.The motion characteristics of quadruped robots can cause motion blur,reducing the reliability of sensor information for subsequent detection tasks.Therefore,a convolutional neural network with an adaptive guidance module is proposed to handle image degradation and improve parameter prediction,thereby improving its performance in detection tasks.Subsequently,the detection results are used as priors for subsequent tasks such as visual navigation and obstacle avoidance.