Distributed dynamic modeling method for multi-leg robot

Abstract

A distributed dynamics modeling method for a multi-leg robot comprises the following steps that (1) based on virtual model control, a trunk single rigid body serving as an object, simplifying legs into virtual actuator input, performing stress analysis, and establishing a single rigid body dynamics model; (2) adopting an iterative leg dynamics model resolving algorithm to establish a leg dynamicsmodel; the leg dynamic model resolving algorithm comprises the following steps: (1) taking a contact point between the tail end of a limb and the environment as a base coordinate system, and extrapolating to obtain an inertia force and a moment borne by the mass center of each rod piece; and (2) starting from the connection point of the trunk and the limbs, taking the force spinor of the limbs acting on the trunk as a controlled quantity, and performing inward pushing to obtain the interaction force spinor of the internal rod piece for realizing the controlled quantity and the contact force spinor of the limbs and the environment. According to the method, the complexity of model establishment is greatly reduced, the dynamics of the multi-leg robot is quickly updated, and the real-time requirement of control is met.

Description

technical field [0001] The invention relates to a dynamic modeling method for a multi-legged robot, belonging to the field of multi-legged robot modeling. Background technique [0002] At present, the motion control methods of multi-legged robots generally include position control, force control and force-position hybrid control. Among them, position control is the most common, based on forward and inverse kinematics combined with compliant control to achieve pre-planned trajectory movement, for example, published in "IEEE / RSJ International Conference on Intelligent Robots and Systems" in 2010 ("IEEE / RSJ International Conference on Intelligent Robots and Systems") "Design and ExperimentalEvaluation of the Hydraulically Actuated Prototype Leg of the HyQ Robot" ("Design and Experimental Validation of the HyQ Robot Hydraulically Actuated Prototype Leg of the HyQ Robot") through the single-leg kinematics analysis, the force Jacobian matrix is ​​obtained, and the position error ...

AI Technical Summary

Problems solved by technology

[0003] Although force control has obvious advantages compared with other control methods, its progress in the application of multi-legged robot control is very slow. Taking the most advanced robot in the world, quadruped robot as an example, as published in "IEEE / RSJ International in 2018 "Dynamic Locomotion in the MIT Cheetah 3 Through ConvexModel-Predictive Control" ("Dynamic Locomotion in the MIT Cheetah 3 Through ConvexModel-Predictive Control") on Conference on Intelligent Robots and System ("IEEE International Conference on Intelligent Robots and Systems") MITcheetah control, and "Planning and Execution of Dynamic Whole-Body Locomotion for a Hydraulic Quadruped on Challenging Terrain" ("Hydraulic Drive Four The whole body dynamic motion planning and execution of foot bionic robot on complex terrain ") used the force control method for the research of ETH HyQ, etc. The force control method used in it has the following two problems: first, the use of Newton's formula and Euler's formula Establish a simplified single rigid body dynamics model of the torso. The inaccuracy of the model leads to the inaccuracy of the expected force screw output of the virtual actuator; second: Lighten the legs so that their mass is negligible compared to the torso Neglected, thus simplifying the calculation of the joint torque to the Jacobian mapping of the desired force screw

In summary, the existing force control methods are based on simplified models and harsh assumptions, and a dynamics-based force control method that can be used for legged robots has not yet been developed.

The above problems are generally caused by inaccurate dynamic models. Traditional dynamic modeling methods such as Lagrangian method and Newton-Euler method are very complicated to model multi-rigid body dynamic systems, and the calculation cost is very high. Too high; the resulting analytical formula is very complex and huge, which does not meet the real-time requirements of control, so it is not conducive to the realization of engineering applications

Images