A kinematic chain is an assembly of rigid bodies (links) connected by joints that constrain their relative motion, forming the structural and mathematical basis for analysing and controlling the motion of mechanisms and robots. Open kinematic chains (serial manipulators) have one free end and exhibit simple forward kinematics but complex inverse kinematics; closed kinematic chains (parallel manipulators) have all links connected in loops, offering higher stiffness and load capacity. The Denavit-Hartenberg convention provides the canonical parameterisation for representing joint geometry and computing transformations along the chain, underpinning all modern robot programming and simulation systems.
Content
- The kinematic chain concept was formalised by Franz Reuleaux in his 1876 work “Theoretische Kinematik,” which introduced systematic analysis of mechanisms as constrained assemblies of links and joints. Reuleaux identified the lower kinematic pair (surface contact, e.g. revolute and prismatic joints) and higher kinematic pair (point or line contact) as the fundamental joint types, and showed that mechanisms are closed kinematic chains where all links are connected. The distinction between open and closed chains was formalised in robotics literature in the 1950s–60s as the first industrial manipulators were developed. Denavit and Hartenberg’s 1955 paper “A Kinematic Notation for Lower-Pair Mechanisms Based on Matrices” established the DH parameter convention—four parameters (link length, link twist, joint offset, joint angle) per joint—that remains the standard representation in robotics textbooks and software.
- Analysis of a kinematic chain proceeds from the chain topology. For an n-DOF serial (open) chain, the forward kinematics problem—computing end-effector position and orientation given joint variables—is solved by composing n homogeneous transformation matrices (4×4 matrices in SE(3)) using the DH convention. The result is unique and computationally straightforward. The inverse kinematics problem—finding joint variables that achieve a desired end-effector pose—is generally non-unique (multiple solutions exist for redundant manipulators with n>6 DOF), may have no solution for poses outside the workspace, and lacks closed-form solutions for general geometries, requiring numerical iterative methods (Jacobian pseudo-inverse, cyclic coordinate descent). Closed kinematic chains (parallel robots such as the Stewart platform and delta robot) provide higher rigidity and load capacity but have complex, constrained forward kinematics that require solving systems of nonlinear equations.
- Kinematic chain analysis is the foundation of all robot simulation, programming, and control software. URDF (Unified Robot Description Format), the standard format for robot model specification in ROS, represents robots as kinematic trees with link geometries and joint parameters. MoveIt, the primary ROS motion planning framework, uses kinematic chain solvers (KDL, TRAC-IK, IKFast) to compute inverse kinematics for arm manipulation tasks. Physics engines including Bullet, MuJoCo (used for reinforcement learning), and NVIDIA Isaac Sim model robots as kinematic chains with constraint solving for contact dynamics. In manufacturing, offline programming (OLP) systems compute optimal joint trajectories for welding, machining, and assembly by solving the kinematic chain along the complete tool path.
- Through 2024–2025, kinematic chain analysis has been extended into machine learning territory. Neural network-based inverse kinematics solvers (trained on sampled configurations) achieve real-time performance for high-DOF systems (humanoid arms with 7+ DOF) that challenge numerical methods. Learning-based approaches also address the calibration problem: identifying the actual DH parameters of a physical robot (which deviate from nominal values due to manufacturing tolerances and wear) by fitting kinematic models to measurement data. Continuum robots—flexible, biologically-inspired manipulators used in minimally invasive surgery and inspection—require Cosserat rod or modal shape function models rather than rigid-link DH parameterisation, extending the kinematic chain concept to infinite-DOF systems described by continuous curvature functions.