Background Research
-
LQR-Assisted whole-body control of a wheeled robot with kinematics loops.
Figure 1: Ascento 2 robot [1]
Figure 2: Ascento 2 dynamic model [1]
The wheel-leg biped robot control and balance are perfectly illustrated by the Ascento 2 robot. The capacity to move quickly and with agility is essential for mobile ground robot deployment. Wheeled-legged systems take advantage of both the efficiency and speed of wheels as well as the strength of legs to navigate obstacles and uneven terrain. Wheeled bipedal robots have begun to demonstrate the characteristics needed for real-world applications in recent years, while also enabling quick and economical designs, requiring fewer actuators and being naturally capable of turning on the spot. [1]
The system dynamics are derived from an explicit formulation of the kinematics, which is straightforward for linkages with basic geometry but significantly more complex for loops with erratic link lengths. [1]
2. A biped wheel-legged robot with improved balancing and disturbance rejection capability assisted by electrical-jets.
In terms of movement, a bipedal wheel-legged robot combines the adaptability of a legged robot with the effectiveness of the wheeled robot when it comes to navigating on flat terrain. As there are only two points of contact between the ground and the wheels of bipedal wheel-legged robots, they present greater difficulties than other types of bipedal robots. Several wheel-legged prototypes with excellent performance are currently on the market, including Handle from Boston Dynamics, Ascento from ETH, Ollie from Tencent, WRL from HIT, and NeZha from SUSTech. Because of limitations imposed by these systems' internal behaviour, they are unable to recover from external impact or disturbance with minimal displacement, and sometimes they even fail to recover at all. [2]
Figure 3: Wheel leg robot prototype [2]
Figure 4: The simulation model in Matlab Simulink with simscapes multibody is shown in A and the corresponding visual display is shown in B. [2]
The construction of a simulation model precedes experimental validation. The foundation of the simulation model is MATLAB Simulink Simscape Multibody. Initially, for the sake of simplification, the parallelogram link between the knee motor and knee joint is treated as being rigidly linked to the thigh because its displacement is quite minimal. In this situation, operating the knee joint directly with a knee motor is equal to installing it there. Secondly, for simplicity of controller design, knee motors and hip motors work in position mode, and the E-Jet is equivalent to a controllable force at the centre of where it is installed. [2]
3. LQR-based Balance control of two-wheeled legged robot.
Figure 5: Conventions and frames of modelling [2]
Wheeled robots are very effective in their execution but less terrain adaptive. A wheeled robot that can do a variety of jobs has been the focus of intense research. But when it comes to adjusting to complex surroundings, wheeled robots are not comparable to legged robots. Bipedal robots have a long history of research on their legs. Currently, this type of robot has advanced to the point that it can function as a humanoid robot, which can adapt to the working environment of humans and successfully resolve various issues. Although the legged robots are less inefficient at moving, they can adapt to the terrain better. [3]
Figure 6: Wheel leg robot prototype [3]
Bipedal and wheeled robots both have advantages, but they work best together. As a result, WLR was created. It has the rapid response and execution accuracy of wheeled robots in addition to the excellent terrain adaptability of legged robots. WLR will have many application scenarios and high economic value if the motion balance problem is successfully solved and artificial intelligence technology is deployed as the brain. In particular, they have significant uses in instructional aids, entertainment films, military equipment, and other areas in schools and universities in addition to being widely employed in airports, hotels, pensions, and other service businesses. Because of its superior adaptability and agility, WLR has a high research value. [3]
Figure 7: Wheel leg robot dynamic model [3]
4. Jumping wheel-legged robot in complex terrain environment.
Figure 8: Schematic diagram of a robot structure [4]
Wheel-legged robots can move quickly and steadily on flat surfaces, but they struggle with poor balance and low mobility levels on uneven surfaces like roadways. The blog post will outline a novel design for a wheel-legged robot with a parallel four-bar mechanism. To achieve stable motion, which will allow the robot to jump over obstacles and adapt to rough terrain, linear quadratic regulator (LQR) and fuzzy proportion differentiation (PD) jumping controllers were designed and developed. [4]
In various topographical situations, simulations and real-world experiments are carried out to confirm the effectiveness of obstacle-crossing techniques. According to the simulation results, when using adaptive retractable wheel-legs to overcome obstacles on terrain with potholes, the maximum height error of the two hip joint motors is 2 mm. However, when using a single leg to overcome obstacles, the maximum height error of the hip joint motors is only 6.6 mm. When simulation data and experimental findings from real-world scenes are compared, it becomes clear that the robot is more robust when navigating challenging terrain. [4]
5. Balance stability augmentation for wheel-legged biped robot through arm acceleration.
Figure 9: Wheel legged robot manipulator arm [5]
The self-balancing wheel-legged robot has a smaller footprint, moves faster, and is more manoeuvrable than legged biped robots. Due to the differential drive's non-holonomic constraints and self-balancing nature, there are still problems that need to be fixed in path planning, motion control, stability analysis, and system modelling. They are more enticing for indoor settings including hotel lobbies, banks, hospitals, restaurants, and warehouses because of these distinctive qualities. There are two types of wheeled inverted pendulum (WIP) robots: those with legs and those without. The lower body was the primary focus of the majority of earlier investigations in this field. [5]
In this paper, a control system is presented to enhance the stability and robustness of an underactuated self-balancing wheel-legged robot. Additionally, we statistically assess the impact of the active arm usage on the robot's improved balance stability in the ROS-Gazebo environment using the centroidal moment pivot (CMP) as a critical indicator. [5]
Figure 10: Lower body dynamic model [5]
6. Modelling and control of a wheel biped robot.
Figure 11: Robot prototype and simplified model [6]
An underactuated nonlinear system like a wheeled biped robot (WBR) makes it challenging to achieve stable control. A decoupled control structure is suggested to enhance a WBR's ability to move dynamically and maintain balance. The WBR is first separated into a five-link multi-rigid body system and a variable-length wheeled inverted pendulum. Then, a time-varying linear quadratic regulator and a model predictive controller are created, respectively, for the two previously mentioned simplified models. The WBR can achieve changing height, resisting external disturbances, velocity tracking, and jumping thanks to the control framework. The outcomes of physical experiments and simulations attest to the framework's efficacy. [6]
While leg-based locomotion is more adaptable and can navigate difficult areas, wheels have the advantages of high efficiency and quick mobility. The best of both worlds can be achieved by combining the benefits of the two. The masses of the individual linkages and their centroid positions are used to weight the position of the analogous centroid. The Denavit-Hartenberg (D-H) convention was utilised to create the kinematic model in order to define the relationship between this centre of mass and the axle coordinate system. [6]
Figure 12: Robot dynamic model [6]
7. Dynamic wheeled motion control of wheel-biped transformable robots.
Figure 13: Robot prototype [7]
The majority of extant biped robots can only move by wheels or by using their feet to walk. This study introduces the dynamic wheeled control of a full-sized wheel-biped transformable robot, SR600-II, comprising wheeled locomotion and in-situ wheel-to-foot (WtF) transformation. It has wheels for moving along level ground and switching modules for assuming a footed stance in the presence of obstacles. The kinematics for wheeled locomotion are first determined under the upper-body lumped centre-of-mass (CoM) constraint. Finally, a wheeled inverted pendulum (WIP) is used to simulate the dynamics of wheeled locomotion by include variables relevant to the upper body's position. The effectiveness of the suggested dynamic wheeled control techniques for both wheeled locomotion and in-situ WtF transformation has been confirmed through simulations and tests on the SR600-II prototype. [7]
Figure 14: Robot dynamic model [7]