The creation of dynamic manipulation behaviors for high degree of freedom, mobile robots will allow them to accomplish increasingly difficult tasks in the field. We are investigating how the coordinated use of the body, legs, and integrated manipulator, on a mobile robot, can improve the strength, velocity, and workspace when handling heavy objects. We envision that such a capability would aid in a search and rescue scenario when clearing obstacles from a path or searching a rubble pile quickly. Manipulating heavy objects is especially challenging because the dynamic forces are high and a legged system must coordinate all its degrees of freedom to accomplish tasks while maintaining balance. To accomplish these types of manipulation tasks, we use trajectory optimization techniques to generate feasible open-loop behaviors for our 28 dof quadruped robot (BigDog) by planning trajectories in a 13 dimensional space. We apply the Covariance Matrix Adaptation (CMA) algorithm to solve for trajectories that optimize task performance while also obeying important constraints such as torque and velocity limits, kinematic limits, and center of pressure location. These open-loop behaviors are then used to generate desired feed-forward body forces and foot step locations, which enable tracking on the robot. Some hardware results for cinderblock throwing are demonstrated on the BigDog quadruped platform augmented with a human-arm-like manipulator. The results are analogous to how a human athlete maximizes distance in the discus event by performing a precise sequence of choreographed steps.

The creation of high degree of freedom dynamic mobile manipulation techniques and behaviors will allow robots to accomplish difficult tasks in the field. We are investigating the use of the body and legs of legged robots to improve the strength, velocity, and workspace of an integrated manipulator to accomplish dynamic manipulation. This is an especially challenging task, as all of the degrees of freedom are active at all times, the dynamic forces generated are high, and the legged system must maintain robust balance throughout the duration of the tasks. To accomplish this goal, we are utilizing trajectory optimization techniques to generate feasible open-loop behaviors for our 28 dof quadruped robot (BigDog) by planning the trajectories in a 13 dimensional space. Covariance Matrix Adaptation techniques are utilized to optimize for several criteria such as payload capability and task completion speed while also obeying constraints such as torque and velocity limits, kinematic limits, and center of pressure location. These open-loop behaviors are then used to generate feed-forward terms, which are subsequently used online to improve tracking and maintain low controller gains. Some initial results on one of our existing balancing quadruped robots with an additional human-arm-like manipulator are demonstrated on robot hardware, including dynamic lifting and throwing of heavy objects 16.5kg cinder blocks, using motions that resemble a human athlete more than typical robotic motions. Increased payload capacity is accomplished through coordinated body motion.

This paper presents a modification for a broad class of controllers based on the LIPM dynamics. We use a change of variables such that instead of controlling the center of mass, we control an “augmented center of mass”, which is unaffected by upper body angular accelerations. We use upper body orientation as an additional source of control authority, allowing us to use both upper body rotation and center of pressure modulation for control. We demonstrate an improved robustness to external pushes with this additional control authority through simulated standing and walking experiments. We also demonstrate the modified controller on our force-controlled humanoid robot.

This paper considers the design of state estimators for dynamic balancing systems using a Linear Inverted Pendulum model with unknown modeling errors such as a center of mass measurement offset or an external force. A variety of process and output models are constructed and compared. For a system containing modeling error, it is shown that a naive estimator (one that doesn't account for this error) will result in inaccurate state estimates. These state estimators are evaluated on a force-controlled humanoid robot for a sinusoidal swaying task and a forward push recovery task.

In order to interact with human environments, humanoid robots require safe and compliant control which can be achieved through force-controlled joints. In this paper, full body step recovery control for robots with force-controlled joints is achieved by adding model-based feed-forward controls. Push Recovery Model Predictive Control (PR-MPC) is presented as a method for generating full-body step recovery motions after a large disturbance. Results are presented from experiments on the Sarcos Primus humanoid robot that uses hydraulic actuators instrumented with force feedback control.

This paper develops full-state parametric controllers for standing balance of humanoid robots in response to impulsive and constant pushes. We also explore a hypothesis that postural feedback gains in standing balance should change with perturbation size. From an engineering point of view this is known as gain scheduling. We use an optimization approach to see if feedback gains should scale with the perturbation for a simulated robot. We simulate models in the sagittal and lateral plane and in 3-dimensions, use a horizontal push of a given size, direction and location as a perturbation, and optimize parametric controllers for different push sizes, directions and locations. During a simulated perturbation experiment, the appropriate controller is continuously selected based on the current push. For an impulse, the simulated robot recovers back to the initial state; for a constant push, the robot moves to an equilibrium position which leans into the push and has zero joint torques. We show the performance of optimized parametric controllers in response to different external pushes.

This paper presents a model-based method, called Dynamic Balance Force Control (DBFC), for determining full body joint torques based on desired COM motion and contact forces for compliant humanoid robots. The center of mass (COM) dynamics are affected directly through contact force control to achieve stable balance. This idea is used to formulate DBFC considering the full rigid-body dynamics of the robot to produce desired contact forces. To achieve generic force control tasks, a virtual model controller, DBFC-VMC, is presented. Results presented from experiments on a force-controlled humanoid robot and simulation demonstrate the general purpose use of this control.

We present work on compliant control of dynamic humanoid balance and walking. We use the linear biped model (LiBM) to model the dynamics of balance on two feet. To achieve periodic motion, as in walking, we derive an orbital energy controller for this model. We also present our methods for applying this control to a torque-controlled humanoid robot, which include estimating the center of mass state and generating feed-forward torque commands.

We combine three threads of research on approximate dynamic programming: sparse random sampling of states, value function and policy approximation using local models, and using local trajectory optimizers to globally optimize a policy and associated value function. Our focus is on finding steady-state policies for deterministic time-invariant discrete time control problems with continuous states and actions often found in robotics. In this paper, we describe our approach and provide initial results on several simulated robotics problems.

We extend simple models previously developed for humanoids to large push recovery. Using these simple models, we develop analytic decision surfaces that are functions of reference points, such as the center of mass and center of pressure, that predict whether or not a fall is inevitable. We explore three strategies for recovery: 1) using ankle torques, 2) moving internal joints, and 3) taking a step. These models can be used in robot controllers or in analysis of human balance and locomotion.

This paper presents a balance controller that allows a humanoid to recover from large disturbances and still maintain an upright posture. Balance is achieved by integral control, which decouples the dynamics and produces smooth torque signals. Simulation shows the controller performs better than other simple balance controllers. Because the controller is inspired by human balance strategies, we compare human motion capture and force plate data to simulation. A model tracking controller is also presented, making it possible to control complex robots using this simple control.

Multiple strategies for standing balance have been observed in humans, including using the ankles to apply torque to the ground, using the hips and/or arms to generate horizontal ground forces, and using the knees and hips to squat. This paper shows that multiple strategies can arise from the same optimization criterion. It is likely that humanoid robots will exhibit the same balance strategies as humans.