3 July 2026 in Informatics and Automation
The purpose of the work is to create methods for controlling real-time walking robots for a wide range of motion modes. A modification of the inverse dynamics method is considered, using a universal adaptive gait generator that implements the six-dimensional motion of the robot's body. This method has been implemented in software, with the addition of regularization methods for numerically integrating equations with a singular matrix. The results of theoretical and experimental research on the motion of a dog robot in various modes over micro and macro irregularities are presented. There is a low level of dynamic parameters of the robot's motion, in particular, high vibration, which corresponds to the literature data for a similar kinematic scheme of dog robots, the control of which is synthesized by machine learning methods and inverse dynamics. The reason for the low quality of control is the phases of motion in which the system becomes uncontrollable. Possible modifications of the robot's kinematic scheme are considered to ensure the unconditional controllability of the system. These modifications can be achieved by introducing redundant controllable degrees of freedom, such as a tail, paws, etc. New theoretical methods for controlling systems with redundant degrees of freedom are proposed based on the generalized dynamic balancing problem. The effectiveness of these methods is confirmed through simulations of the controlled movement of a dog robot with a tail and paws. For a common kinematic scheme of a dog robot with three drives per leg and point contact in the foot, a rational solution is to use a single wheel for moving over micro-irregularities and walking propulsors for overcoming macro-irregularities and escaping from emergency situations, such as tipping over. The episodic use of walking propulsors allows for simplified control modes, which are effectively implemented within the proposed modification of the inverse dynamics method. The concept has been confirmed by numerical and field experiments. In this case, the locomotion control system can be implemented on an inexpensive hardware basis, including the use of proven instrumentation for widespread robots, such as quadrocopters. A distinctive feature of robots based on the proposed control principles is their modular design, which allows for modifications in a wide range of technical characteristics through various module configurations. Such highly mobile robots can become a cheap and safe platform for refinement of navigation systems, unmanned control, etc. in real-world conditions in various fields of application.