Open Source Robotics

A curated collection of high-quality open source robotics projects, tools, and software to propel the robotics community forward.

Single Rigid Body Model

  • Also called: SRB model, SRBD, single rigid body dynamics
  • Category: Control › Locomotion control
  • Tags: legged-locomotion, reduced-order-models, model-predictive-control, quadrupeds, rigid-body-dynamics

A reduced-order model of a legged robot as one rigid body with fixed mass and inertia, moved by contact forces at massless legs; widely used in locomotion MPC.

Definition

The single rigid body (SRB) model represents a legged robot as one rigid body that carries the robot's whole mass and a rotational inertia, usually taken as constant in the body frame. The legs are treated as massless, so their only role is to transmit ground reaction forces from the feet to the body. The state is the body's position, orientation, linear velocity and angular velocity, and the inputs are the contact forces at the feet in stance.

The dynamics are the Newton-Euler equations. The sum of contact forces plus gravity equals mass times the acceleration of the center of mass, and the sum of the moments of the contact forces about the center of mass, each the cross product of the foot position relative to the center of mass with its force, equals the rate of change of angular momentum. External forces and moments, such as a load pulled by the robot or forces from another robot it is attached to, can be added to these sums directly. The model is close to centroidal dynamics, which also describes the robot through its center of mass and total momentum, but the SRB model ignores how leg motion changes the inertia and contributes angular momentum.

The model became a standard basis for model predictive control of quadrupeds. A widely used formulation for the MIT Cheetah 3 linearized the rotational dynamics by assuming small roll and pitch and a known yaw trajectory, which turns the search for contact forces over a short horizon into a convex quadratic program with friction cone constraints. Nonlinear versions keep the full rotation dynamics. The SRB model works best when the legs are light compared with the body; for humanoids with heavy limbs, or motions such as flips where leg swing matters, centroidal or whole-body models are more accurate.

Why it matters

The SRB model is simple enough to optimize over a horizon of several steps in real time on robot hardware, yet it keeps the body's 3D rotation and the moment created by each foot force. That balance has made it a common choice for force-based locomotion controllers on quadrupeds and for planners that must account for external loads.

Key applications

  • Model predictive control of quadruped locomotion
  • Distributing ground reaction forces among stance feet
  • Planning jumps and dynamic maneuvers that involve body rotation
  • Collaborative locomotion with external loads or connected robots
  • Gait and contact schedule optimization

In practice

On a quadruped, a model predictive controller built on the SRB model runs at a fixed rate. At each update it takes the estimated body state, the contact schedule from the gait and the desired velocity, and solves for the ground reaction forces of the stance feet over the coming horizon, subject to friction cone constraints and force limits. The first set of forces is converted to joint torques with the transpose of each leg's Jacobian, while swing legs follow foot trajectories under a separate controller. When the robot carries a payload or is connected to other robots, the expected external force and its moment about the center of mass are added to the model so that the planned foot forces compensate for them.

Common misconceptions

  • The SRB model ignores body rotation — it keeps full 3D orientation and angular velocity; only leg mass and leg-induced momentum are dropped
  • The SRB model and centroidal dynamics are the same — centroidal dynamics accounts for how limb motion changes the total inertia and momentum, which the SRB model fixes or ignores
  • Convex SRB MPC is valid for any motion — the common convex form relies on small roll and pitch, so large rotations such as flips need nonlinear or rotation-aware formulations

Who should know this

  • Quadruped and humanoid locomotion engineers
  • Model predictive control developers
  • Researchers on legged robots carrying loads or working in teams
  • Students learning reduced-order models of legged robots

Related technologies

  • Convex Model Predictive Control
  • Newton-Euler Equations
  • Centroidal Dynamics
  • Quadratic Programming
  • Whole-Body Control

Related terms

  • Centroidal Dynamics
  • Model Predictive Control
  • Friction Cone
  • Quadruped
  • Whole-Body Control
  • Linear Inverted Pendulum

Related projects

  • DimOS: Python robotics runtime with agents, SLAM and teleop, no ROS required
  • MuJoCo MPC: DeepMind's interactive real-time model predictive control framework for MuJoCo
  • PACE Sim2Real: Systematic sim-to-real transfer via actuator dynamics identification
  • python_simple_mppi: MPPI control derived and implemented in numpy, with four worked examples
  • mjbatch: Runs thousands of MuJoCo simulations in parallel on the CPU
  • Quadruped-PyMPC: Python MPC framework for quadruped locomotion with gradient and sampling control

This website requires JavaScript to function properly. Please enable JavaScript to explore our collection of open source robotics projects.