Development of a new underwater gliding snake robot!

With the growing demand for exploration in complex underwater environments, higher requirements are placed on the endurance maneuverability of corresponding detection equipment. The underwater robot research development glider has the advantages of low power consumption, low noise long endurance, is suitable for wide water environments such as rivers oceans;

However, its fixed structure reliance on net buoyancy for propulsion result in poor maneuverability of underwater gliders, making it difficult to effectively turn avoid obstacles. Snake-shaped fish-shaped bionic underwater robots have the characteristics of high maneuverability can be flexibly applied to complex underwater environment tasks such as underwater building crack detection, underwater equipment maintenance underwater rescue, but they have the disadvantages of low energy efficiency poor endurance.

The underwater snake-like robot developed by eelume. Therefore, the Shenyang Institute of Automation of the Chinese Academy of Sciences has developed an underwater gliding snake-like robot that combines the characteristics of an underwater glider an underwater snake-like robot. The robot can realize gliding motion driven by net buoyancy a variety of swimming gaits driven by joint torque, has the advantages of strong endurance strong maneuverability.

A researcher the Chinese Academy of Sciences said: "In response to the problem of difficult modeling of hydrodynamic environments, we used reinforcement learning methods to make the underwater gliding snake robot adapt to complex water environments automatically learn to control gliding motion by adjusting buoyancy alone. In this regard, we proposed a recurrent neural network Monte Carlo policy gradient algorithm to improve the problem of difficulty in training the algorithm due to the difficulty in fully observing the robot's state."

We also approximate the basic gliding motion control problem of the underwater gliding snake robot as a Markov decision process, thereby obtaining an effective gliding control strategy, verify the effectiveness of the proposed method through simulation experiments.

The gliding motion system of the underwater gliding snake-like robot has system nonlinear coupling; disturbances caused by actual fluid uncertainty, sensor measurement noise, actuator error, etc. may affect the tracking effect of the system; some state quantities (such as gliding speed) are difficult to measure directly through sensors, the concept of a nonlinear closed-loop control system framework is proposed.

Through decoupling feedback linearization, the complex nonlinear system is transformed into an equivalent linear form, then a sliding mode controller is designed based on the reaching law method to reduce jitter achieve robust control. The measurement noise is filtered the unknown state quantity is estimated through the unscented Kalman filter. This method handles the nonlinear transfer problem of the mean covariance through unscented transformation, retains the high-order terms of the system, has the advantages of high accuracy strong stability. The proposed closed-loop control framework has certain versatility for some other nonlinear systems.

The underwater gliding snake-like robot achieves a perfect combination of long range maneuverability. The self-expanding part enables the underwater gliding robot to glide underwater with low power. The snake-like structure enables the robot to be flexible maneuverable underwater, it will be flexibly used in various complex underwater environment tasks.