Skip to main content

Advanced Control Strategies

Overview​

While ZMP-based control and the LIPM are powerful tools for generating stable walking patterns, they are not sufficient for more dynamic and complex behaviors. This section introduces some of the advanced control strategies that are used in modern humanoid robots, including Model Predictive Control (MPC), whole-body control, and compliance and force control.

Detailed Explanation​

Model Predictive Control (MPC)​

Model Predictive Control (MPC) is an advanced control strategy that uses a model of the robot to predict its future behavior. The controller then optimizes the robot's control inputs (i.e., the joint torques) over a short time horizon to achieve a desired goal, such as tracking a desired trajectory or maintaining balance.

One of the key advantages of MPC is that it can handle constraints, such as the limits on joint torques and the need to keep the robot's feet on the ground. This makes it well-suited for controlling a complex system like a humanoid robot.

The basic steps of MPC are:

  1. At each time step, the controller uses a model of the robot to predict its future behavior over a short time horizon.
  2. The controller then solves an optimization problem to find the sequence of control inputs that will minimize a cost function. The cost function typically includes terms that penalize deviations from a desired trajectory and the amount of control effort used.
  3. The first control input in the sequence is applied to the robot.
  4. The process is repeated at the next time step.

Whole-Body Control​

Whole-body control is a control strategy that considers the full dynamics of the robot. Unlike simpler control strategies that treat the arms, legs, and torso as separate systems, whole-body control coordinates the motion of all the joints in the robot to achieve a common goal.

This approach is particularly important for tasks that require the robot to use its whole body, such as lifting a heavy object or pushing against a wall. By coordinating the motion of all the joints, the robot can distribute the load more effectively and maintain its balance.

Compliance and Force Control​

Compliance is the ability of a robot to deform in response to an external force. Force control is the ability of a robot to actively control the forces that it applies to its environment. Both are essential for safe and effective human-robot interaction.

A compliant robot is less likely to damage itself or its environment if it collides with an object. It can also be more effective at tasks that require physical contact with the environment, such as cleaning a window or assembling a product.

There are two main approaches to achieving compliance in a robot:

  • Passive compliance: This is achieved through the use of springs or other compliant elements in the robot's joints.
  • Active compliance: This is achieved by using a control system to actively modulate the stiffness of the robot's joints. This can be done by using a force sensor to measure the contact forces and then adjusting the joint torques accordingly.

Hands-on Exercise: Designing a Whole-Body Controller​

This exercise is a conceptual task to get you thinking about the challenges of whole-body control.

The Scenario​

You are tasked with designing a whole-body controller for a humanoid robot that needs to open a heavy door. The robot has two arms and two legs, and it is equipped with force sensors in its hands and feet.

Your Task​

  1. Describe the different sub-tasks that are involved in opening the door. For example, the robot needs to walk to the door, reach for the handle, grasp the handle, turn the handle, and pull the door open.
  2. For each sub-task, identify the main control objective. For example, for the walking sub-task, the main objective is to maintain balance while moving towards the door.
  3. For each sub-task, identify the key constraints that the controller needs to satisfy. For example, when pulling the door open, the controller needs to make sure that the forces applied to the feet do not exceed the friction limit (to prevent slipping).
  4. Describe how you would use a hierarchical control approach to coordinate the different sub-tasks. For example, you could have a high-level planner that decides which sub-task to execute, and a set of mid-level controllers that are responsible for executing each sub-task.
  5. Explain how you would use compliance and force control to make the robot more robust to uncertainties in the environment. For example, how would you use force control to prevent the robot from pulling too hard on the door handle?