Objective: A two-legged wheeled robot is a new composite ground mobile robot with wheels designed at the end of the legs. The design combines the improved maneuverability and flexibility of traditional wheeled mobile platforms with footed robotic structures and has improved application scenarios and research value. However
motion control of two-legged wheeled robots depends highly on the accurate dynamics model. At the same time
it has control difficulties such as underdrive
strong coupling
and nonlinearity
leading to problems in achieving motion and self-balancing control. Therefore
proposing an efficient control method for the motion control of two-legged wheeled robots is practically significant. Methods: The overall dynamics model of the two-legged wheeled robot has a large structure
which is not conducive to the controller characterization and construction. Thus
this paper adopts a distributed model to establish dynamics models of the leg-wheel and torso subsystems. This approach retains all the dynamics characteristics of the robot and connects them via the inter-module motion/force transfer relationship to complete the whole-body dynamics of the wheel-leg-torso interaction force as the output. The joint moment solver is constructed with the end output force as the task space based on this model. TFeed-forward compensation of the rod inertia force caused by the initial state quantity is performed using the observation signal. Then
a distributed control framework with the torso position as the task space is proposed to hierarchically plan the torso joint force and moment. The walking motion control architecture based on the whole-body moment is constructed
as shown in Fig. 5. An adaptive planning method for the longitudinal trajectory of the torso while jumping is established based on the walking motion control architecture. This approach gives the torso the longitudinal velocity required for jumping by planning the longitudinal motion. A cubic polynomial plans the torso height trajectory of the jumping support phase to ensure the smoothness of the velocity and acceleration. The torso height trajectory of the jumping airborne phase is obtained by integrating the velocities. Variations in its torso height trajectory are shown in Eq. 13. A control method of the airborne phase is proposed based on the virtual model and momentum moment theorem. The robot enters a free-fall state after jumping when the torso reaches the jumping speed by controlling leg contraction to remove contact between the wheels and the ground. The entire jumping process is illustrated in Fig. 4. Finally
the effectiveness of the whole-body moment control system is proven via simulations and prototype testing. Results and Discussions: A simulation platform for the two-legged wheeled robot is developed in MATLAB to prove the feasibility of the distributed dynamics modeling and whole-body moment control. Two walking scenarios (linear and circular motion) are established
with three jumping simulations at different heights. The velocity curves for linear motion are shown in Fig. 8
with a maximum error of 0.09 m/s. The robot’s pitch angle and height curves are given in Fig. 9
indicating height errors ranging from -11–2 mm and pitch angle errors ranging from -1–6 mrad. The velocity curves for circular motion are displayed in Fig.13
with a maximum error of 0.071 m/s. The pitch angle and height curves are shown in Fig. 14
indicating a maximum height error of 8.633 mm and a maximum pitch angle error of 6 mrad. The three jumping heights are 55
50
and 45 cm. The jumping results are summarized in Table 1
indicating a maximum jumping height error of 20 mm. This confirms the feasibility of the proposed distributed dynamics modeling and whole-body moment control framework. A prototype of the two-legged wheeled robot is developed
and multiple motion mode experiments are performed. The prototype’s pitch angle and height curves are displayed in Fig. 22
indicating a maximum trunk height error of 3.38 mm and a maximum pitch angle error of 40 mrad. Comparing and analyzing the simulation results and prototype experiments suggests that coordinated wheel and leg movement can achieve extended motion with greater accuracy while maintaining dynamic balance using the trunk as the task space. This confirms the correctness of the proposed distributed dynamics modeling and whole-body moment control framework. This work provides a practical reference for studying the motion control of two-legged wheeled robots. Conclusions: This paper proposes a distributed whole-body dynamics modeling method to preserve the dynamic characteristics of two-legged wheeled robots. The results establish dynamics models of the torso and leg-wheel subsystems while mapping the robot’s joint moments to the end output force. A whole-body moment control framework is proposed based on the distributed dynamics model to achieve motion control and dynamic balance. The framework controls two-legged wheeled robots with inherent instability
strong coupling
and nonlinearity. A multi-motion mode planning method is proposed based on the whole-body moment control framework to improve motion performance. Finally
a simulation platform is built for jumping and walking motion modes
verifying the feasibility of the whole-body moment control system based on the distributed dynamics model. A prototype of a two-legged wheeled robot using synchronous belt-driven tandem legs and a carbon fiber plate frame is developed and tested in multiple motion modes. The results prove the correctness of the distributed dynamics model and the whole-body moment control framework
原文链接:https://jsuese.scu.edu.cn/zh/article/doi/10.12454/j.jsuese.202300854/