A Compliant Hybrid Zero Dynamics Controller for Stable, Efficient and Fast Bipedal Walking on MABEL
read more
Citations
Control strategies for active lower extremity prosthetics and orthotics: a review
Rapidly Exponentially Stabilizing Control Lyapunov Functions and Hybrid Zero Dynamics
Design Principles for Energy-Efficient Legged Locomotion and Implementation on the MIT Cheetah Robot
Models, feedback control, and open problems of 3D bipedal robotic walking
Variable Stiffness Actuators: Review on Design and Components
References
Nonlinear Control Systems
A new model for control of systems with friction
Legged Robots That Balance
Zero-moment point — thirty five years of its life
Related Papers (5)
Frequently Asked Questions (15)
Q2. What future works have the authors mentioned in the paper "A compliant hybrid zero dynamics controller for stable, efficient and fast bipedal walking on mabel" ?
Future research will be directed towards obtaining analytical and experimental results in these areas.
Q3. How much force is required to achieve similar speed gains and speed drops?
In order to achieve similar speed gains and speed drops, the required force from simulation is around 78 N in the forward direction, and around 71 N in the reverse direction, respectively.
Q4. Why is the torso selected as a controlled variable instead of the stance leg angle?
The torso is selected as a controlled variable instead of the stance leg angle, because, for MABEL, the torso represents over 65% of the mass of the robot, and hence the position of the torso heavily influences the gait.
Q5. What is the configuration space of the unconstrained dynamics of the robot?
The configuration space Qe of the unconstrained dynamics of MABEL is a simply-connected subset of S7 × R2: five DOF are associated with the links in the robot’s body, two DOF are associated with the springs in series with the two leg-shape motors, and two DOF are associated with the horizontal and vertical position of the robot in the sagittal plane.
Q6. How was it possible to implement the virtual constraints through a simple PD controller?
For RABBIT, it was possible to implement the virtual constraints through a simple PD controller (Westervelt et al., 2004), peru = −KP y −KDẏfor y given by (22), and ẏ computed numerically.
Q7. Why are the torques discontinuous at phase boundaries?
The torques are discontinuous at phase boundaries, as noted earlier, due to the choice of the virtual constraints being C1 at phase boundaries.
Q8. What is the purpose of the feedback in the optimization process of gait design?
It is used in the optimization process of gait design in order to evaluate the torques along a solution of the model respecting the virtual constraints.
Q9. What are the constraints that need to be satisfied at impact?
Then the generalized external impulsive force acting on the system is obtained from the principle of virtual work as,Fext =(∂ptoesw ∂qe)TIR +(∂qBspst ∂qe)TτR. (15)The authors have three constraints that need to be satisfied at impact.
Q10. What are the criticisms of zero dynamics controllers?
Zero dynamics controllers are often criticized for being overly dependent on the model being accurate, and for being too complex to implement in real time.
Q11. Why is the hip position chosen as an independent coordinate instead of the center of mass?
The hip position is chosen as an independent coordinate instead of the center of mass because it was observed that this choice significantly reduces the number of terms in the symbolic expressions for the dynamics.
Q12. Why did the spring not decompress to the 5 trigger point?
The spring was not decompressing to the 5◦ trigger point, and was probably due to the initial few steps being far away from the nominal orbit, and also because of inability of the controller to accurately track the stance motor leg shape virtual constraint.
Q13. What is the stability of the fixed-points with the proposed closed-loop controller?
The stability of the fixed-points with the proposed closed-loop controller (43) can be tested numerically using a Poincaré map P : S → S with the switching surface taken to be the switching surface at the si → sd event transition, i.e., S = Ssi→sd, andP (xs) = φ (TI ◦∆si→sd (xs) ,∆si→sd (xs)) , (44)where, φ (t, x0) denotes the maximal solution of (12), with initial condition x0 at time t0 = 0 and with u as defined in (43).
Q14. What is the way to improve the speed of a biped?
The speed of a biped can be enhanced by careful mechanism and control design as suggested in (Koechling and Raibert, 1993), and demonstrated in robots such as RunBot (Manoonpong et al., 2007).
Q15. How is the walking gait obtained by optimizing?
The analysis shows that the walking gait obtained by optimizing (39) and with the closed-loop controller (43) is exponentially stable with a dominant eigenvalue of 0.6921.