Friction Cone

The friction cone is a right circular cone whose axis is aligned with the contact normal vector. The cone's half-angle (φ) is defined by the friction coefficient (μ) as φ = arctan(μ). Any resultant friction force vector whose tip lies within this cone represents a state where the contact remains in static friction (no sliding). A force vector lying on the cone's surface indicates impending slip, where the friction force is at its maximum permissible magnitude.

The cone's shape directly encodes the Coulomb friction model, which states:

• Static Friction: |f_t| ≤ μ * |f_n|. The tangential force magnitude is less than or equal to the coefficient of friction times the normal force.
• Kinetic Friction: When sliding occurs, |f_t| = μ_k * |f_n|, where μ_k is the kinetic friction coefficient (often slightly less than the static μ). The cone's interior satisfies the inequality, while its boundary satisfies the equality. This model is independent of apparent contact area and sliding velocity in its basic form.

For computational efficiency in constraint solvers (like those solving a Linear Complementarity Problem), the friction cone is often approximated as a polyhedral pyramid. Common approximations use 4 or 8 edges. This linearization transforms the nonlinear cone constraint into a set of linear inequalities, allowing the use of efficient linear programming or LCP solvers. The trade-off is a slight over-approximation of the feasible force space, which can make the simulated contact slightly more 'sticky' than the true cone.

The standard friction cone assumes isotropic friction, meaning the friction coefficient is identical in all directions tangential to the contact. Anisotropic friction occurs when the friction coefficient differs with direction (e.g., brushed metal, wood grain). This is represented by an elliptical or non-circular cone, where the half-angle φ varies with the direction in the tangent plane. Simulating this requires a more complex model, often defined by two principal friction coefficients (μ_x, μ_y) aligned with a material's anisotropic axes.

In a physics engine's constraint solver, the friction cone defines the feasible set for contact force solutions. The solver's job is to find a set of normal and friction forces that:

• Satisfy non-penetration (normal force is repulsive).
• Lie within the friction cone (or its polyhedral approximation).
• Obey Newton's laws for the entire multi-body system. This is typically formulated as a complementarity problem: The relative tangential velocity at the contact must be complementary to the friction force, meaning sliding only occurs when the force is on the cone's boundary.

For simulation debugging, the friction cone is visualized as a wireframe cone emanating from the contact point. This allows engineers to:

• Verify the calculated friction force vector against the cone boundary.
• Diagnose unstable simulation or unrealistic sliding, which may appear as force vectors outside the cone.
• Tune friction coefficients and solver parameters by observing how the cone size and force solutions interact. This visual tool is critical for system identification and validating that simulated contact behavior matches real-world physics expectations.

In robotic grasping, the friction cone determines if a planned grip will be stable. For a multi-fingered grasp to resist external wrenches (forces and torques), the resultant contact force at each fingertip must lie within its friction cone. This is formalized by the Grasp Matrix and Grasp Wrench Space. A grasp is force-closure if any external wrench can be balanced by contact forces lying strictly within their respective cones.

• Key Calculation: The friction coefficient between the gripper material (e.g., silicone, rubber) and the object surface defines the cone's half-angle.
• Example: A parallel-jaw gripper picking up a box must apply a normal force sufficient that the required tangential force to support the weight lies within the cone.

For legged robots like quadrupeds or humanoids, the friction cone dictates stable foot placement and gait. The Ground Reaction Force (GRF) at each foot must remain inside the cone to prevent slipping. Motion planners and controllers use this constraint to compute feasible footholds and force distributions.

• Zero Moment Point (ZMP): The classic ZMP criterion for bipedal walking implicitly assumes infinite friction. For dynamic motions or low-friction surfaces, explicit friction cone constraints are added to the quadratic program solved by the controller.
• Terrain Adaptation: On slopes or loose terrain, the effective cone angle changes, requiring the robot to adjust its center of mass trajectory and footfall timing.

Tasks like pushing, sliding, or pivoting an object across a surface are governed by friction cone constraints. Quasistatic planners model the limit surface—the set of all possible friction wrenches—which is derived from integrating friction cones over the entire contact area.

• Pushing Mechanics: To predict if a push will cause sliding or rotation, the line of action of the pusher's force is checked against the object's composite friction cone and limit surface.
• Non-Prehensile Manipulation: This is essential for manipulating objects too large or heavy to grasp, where the robot uses controlled slipping and re-establishing contact.

Model Predictive Control (MPC) and trajectory optimization frameworks for robots explicitly include friction cone constraints as inequality constraints in their numerical solvers. This ensures the optimized motion plans are dynamically feasible.

• Mathematical Form: The constraint is often linearized, requiring the tangential force components (f_t) to satisfy (\sqrt{f_{tx}^2 + f_{ty}^2} \leq \mu f_n), where (\mu) is the friction coefficient and (f_n) is the normal force. This is commonly approximated as a pyramidal approximation (linear constraints) for speed.
• Solvers: These constraints are handled by Quadratic Programming (QP) or Nonlinear Programming (NLP) solvers like SNOPT or IPOPT.

Physics engines (e.g., MuJoCo, Bullet, Drake) implement the friction cone within their constraint solvers. The Linear Complementarity Problem (LCP) or Convex Optimization formulation solves for contact impulses that satisfy Coulomb's law and prevent inter-penetration.

• Pyramidal vs. Conical: For computational efficiency, many real-time simulators use a pyramidal approximation of the cone (e.g., a 4-sided pyramid). High-fidelity off-line simulators may use a true elliptical or conical model.
• Anisotropic Friction: Some engines support direction-dependent friction coefficients, resulting in an elliptical friction cone rather than a circular one.

Precision assembly requires managing frictional forces during insertion. The friction cone defines the allowable contact forces between the peg and hole chamfer to guide the peg without jamming.

• Jamming Condition: Jamming occurs when the resultant contact force lies outside the friction cone, causing wedging. Force-guided strategies actively control the insertion force to keep it within the cone.
• Compliant Motion: Strategies like Remote Center of Compliance (RCC) passively obey friction cone constraints through mechanical design, allowing the peg to align during insertion despite contact forces.

Coulomb friction is the classical dry friction model that underpins the friction cone. It defines two regimes:

• Static Friction: The maximum tangential force that can be exerted before sliding begins, proportional to the normal force.
• Kinetic Friction: A constant, slightly lower tangential force opposing motion during sliding. The model's key simplification is that the friction force magnitude is independent of the apparent contact area, depending only on the normal force and the material-dependent coefficient of friction.

A contact force is the resultant force vector generated at the interface between two colliding rigid bodies. In simulation, it is decomposed for analysis and computation:

• Normal Force: Acts perpendicular to the contact surface, preventing inter-penetration. Its magnitude is determined by material compliance and solver constraints.
• Tangential (Friction) Force: Acts parallel to the contact surface, opposing relative motion. Its allowable direction and magnitude are defined by the friction cone. Accurate modeling of this decomposition is critical for stable, realistic simulation of sliding, rolling, and toppling behaviors.

The Linear Complementarity Problem is a core mathematical formulation used by physics engines to solve contact and friction constraints simultaneously. For the friction cone, it enforces the complementarity conditions:

• The normal force must be non-adhesive (≥0).
• Bodies either have zero penetration (if separated) or a positive normal force (if in contact), but not both—they are complementary.
• The tangential friction force must lie within the cone and be complementary to the relative sliding velocity. Solving the LCP provides the exact impulses or forces that satisfy all contact laws without penetration.

The constraint solver is the algorithmic core of a physics engine that iteratively calculates the forces or impulses needed to satisfy all system constraints, including contact non-penetration and friction cones. Modern solvers like Sequential Impulse or Projected Gauss-Seidel:

1. Receive a set of potential contacts from the collision detection phase.
2. For each contact, formulate constraints based on the friction cone and restitution.
3. Solve the resulting system (often as an LCP) over multiple iterations to converge on a physically plausible set of contact forces that prevent drift and jitter.

The Coefficient of Restitution is a scalar value (0 to 1) that models the elasticity of a collision. While the friction cone governs tangential forces, the COR governs the normal direction's energy loss:

• COR = 1: Perfectly elastic collision (e.g., superball). Relative velocity reverses perfectly.
• COR = 0: Perfectly inelastic collision (e.g., clay). Bodies stick together with no rebound. In simulation, the COR scales the relative normal velocity at the contact point post-collision. It is applied alongside friction constraints in the impulse calculation.

A rigid body is the fundamental simulated entity in classical dynamics. It is an idealized solid object where the distance between any two points remains constant, regardless of applied forces. Its state is fully described by:

• Position & Orientation (in world space)
• Linear & Angular Velocity
• Mass & Inertia Tensor The friction cone applies at the contact points between these bodies. The assumption of perfect rigidity simplifies force computation by ignoring internal deformation, focusing all contact interaction at discrete points or surfaces.