Robot Kinematics

Learning Objectives

By the end of this page, you should be able to:

• Distinguish forward kinematics from inverse kinematics and state what each one solves for
• Identify degrees of freedom, joint types, and link parameters in a simple kinematic chain
• Compute the end-effector position of a simple 2-link planar arm using forward kinematics
• Explain why inverse kinematics can have zero, one, or multiple solutions
• Define workspace and singularity, and explain why both matter for real robot design

Quick Answer

Robot kinematics is the study of a robot's motion — position, orientation, and velocity of its links — purely in terms of geometry, without considering the forces or torques that cause the motion. It answers two complementary questions: forward kinematics asks "given these joint angles, where is the end-effector?" and inverse kinematics asks "given a target end-effector position, what joint angles get me there?" Kinematics matters because before you can control a robot arm to pick up an object, you must first know where its tip actually is for a given set of joint values, and what joint values will place its tip at a desired target. Every robot arm controller, from a 3D printer to a surgical robot, relies on kinematic equations running continuously in the background.

Building Blocks: Links, Joints, and Degrees of Freedom

A robot arm is a kinematic chain — a sequence of rigid links connected by joints. Each joint allows one specific type of relative motion:

• Revolute joint — rotation about a single axis (like an elbow). Its variable is an angle, θ.
• Prismatic joint — sliding translation along a single axis (like a piston). Its variable is a displacement, d.
• Spherical/planar joints — combinations of the above, used less commonly in simple arms.

A degree of freedom (DOF) is one independent joint variable. A robot arm with three revolute joints has 3 DOF; each joint contributes exactly one independent way the arm can move. In general, to place an end-effector at any position and orientation in 3D space you need 6 DOF (3 for position, 3 for orientation) — this is why most industrial arms have exactly six joints.

Why it matters: DOF count directly tells you what tasks a robot can and cannot do. A 3-DOF arm can reach a target point but usually can't also control the orientation of its gripper independently — you need at least 6 DOF for full pose control.

Forward Kinematics: From Joint Angles to Position

Forward kinematics answers: given known joint values, where is the end-effector? This is always a well-defined, unique calculation — you plug in numbers and get one answer.

Worked example: consider a simple 2-link planar arm lying flat, with link lengths l₁ and l₂, and joint angles θ₁ (at the base) and θ₂ (at the elbow, measured relative to link 1).

The position of the elbow joint is:

x₁ = l₁ cos(θ₁) y₁ = l₁ sin(θ₁)

The position of the end-effector (adding link 2, whose angle relative to the ground is θ₁ + θ₂) is:

x = l₁ cos(θ₁) + l₂ cos(θ₁ + θ₂) y = l₁ sin(θ₁) + l₂ sin(θ₁ + θ₂)

Plug in θ₁ = 30°, θ₂ = 45°, l₁ = l₂ = 1 m, and you get one specific (x, y) — no ambiguity. This is the essence of forward kinematics: straightforward trigonometry chained link by link.

For arms with more links and 3D motion, engineers use a standardized bookkeeping method called Denavit-Hartenberg (D-H) parameters, which describe each link's geometry with just four numbers (link length, link twist, link offset, joint angle) and combine them via transformation matrices. The math is more involved, but the underlying idea is identical: chain each link's local transformation onto the previous one to find where the end-effector ends up.

Real-world example: a 3D printer's gantry uses (trivial, since it's Cartesian) forward kinematics constantly — every time it moves the print head, its controller is computing where the nozzle tip is based on motor step counts.

Inverse Kinematics: From Target Position to Joint Angles

Inverse kinematics flips the question: given a desired end-effector position (and often orientation), what joint values achieve it? This is the harder, more interesting problem, because unlike forward kinematics, it is not always uniquely solvable.

For the same 2-link arm, given a target (xd, yd), you can solve for θ₂ using the law of cosines:

cos(θ₂) = (xd² + yd² − l₁² − l₂²) / (2 l₁ l₂)

Notice immediately that this equation can have:

• No solution — if the right-hand side is outside [−1, 1], the target is physically out of reach.
• Two solutions — because cos(θ₂) = cos(−θ₂), there are generally two elbow configurations ("elbow up" and "elbow down") that reach the same point.
• One solution — only at the boundary of the workspace, where the arm is fully extended.

This is the single most important conceptual difference between forward and inverse kinematics: forward kinematics is a straightforward function evaluation, while inverse kinematics requires solving equations that may be non-unique, nonlinear, or have no solution at all.

Why it matters: a robot controller doing pick-and-place needs inverse kinematics to convert a vision system's "the object is at this (x, y, z)" into "move these joints to these angles." Getting the wrong one of two valid elbow configurations can send the arm into a collision with itself or the workspace.

Workspace and Singularities

The workspace is the complete set of positions (and orientations) the end-effector can physically reach, given the joint ranges and link lengths. It's shaped by the number of DOF, the joint limits, and any physical obstructions. A robot with more DOF and longer links generally has a larger, more flexible workspace, but reach isn't free — longer links reduce structural stiffness and precision.

A singularity is a special configuration where the robot loses at least one degree of freedom of motion — small changes in joint velocity can no longer produce motion in some direction, or the inverse kinematics solution becomes undefined/infinite. The classic example: when a 2-link arm is fully extended (θ₂ = 0), any attempt to move directly "outward" along the arm requires unrealistically large joint velocities, because the arm is already at the boundary of its reach.

Common misunderstanding: students often think singularities are rare edge cases that can be ignored. In practice, robot motion planners must actively avoid singularities during a trajectory, because near a singularity, small errors in the target position translate into huge, sometimes physically impossible, required joint velocities — this can cause a real robot to jerk violently or stall.

Key Terms

Term	Definition
Kinematic chain	A series of rigid links connected by joints
Degree of freedom (DOF)	One independent joint variable that contributes to the robot's possible motion
Revolute joint	A joint allowing rotation about one axis, described by an angle θ
Prismatic joint	A joint allowing translation along one axis, described by a displacement d
Forward kinematics	Computing end-effector position/orientation from known joint values
Inverse kinematics	Computing joint values required to achieve a desired end-effector position/orientation
Denavit-Hartenberg (D-H) parameters	A standardized 4-parameter method for describing link geometry in a kinematic chain
Workspace	The set of all positions/orientations an end-effector can physically reach
Singularity	A configuration where the robot loses a degree of freedom of achievable motion

Common Mistakes

1. Misconception: "Forward and inverse kinematics are basically the same calculation done in different directions." Why it's wrong: forward kinematics is a direct function evaluation with exactly one answer. Inverse kinematics involves solving (often nonlinear, trigonometric) equations that may have zero, one, or multiple valid solutions. Correct understanding: forward kinematics is always well-posed; inverse kinematics requires checking reachability and choosing among possible configurations (like elbow-up vs. elbow-down).

2. Misconception: "More degrees of freedom always make a robot better." Why it's wrong: extra DOF increase mechanical complexity, cost, and the difficulty of inverse kinematics (more possible solutions, more chances for self-collision), without necessarily improving task performance if the task doesn't need that flexibility. Correct understanding: the right DOF count is the minimum needed for the task — 6 DOF for full pose control in 3D, fewer for simpler planar tasks, more (redundant, 7+ DOF) only when obstacle avoidance or flexibility genuinely requires it.

3. Misconception: "Kinematics tells you how a robot will actually move when you send a command." Why it's wrong: kinematics is purely geometric — it ignores mass, inertia, motor torque limits, and friction. A kinematically valid trajectory can still be dynamically infeasible (too fast for the motors to achieve). Correct understanding: kinematics answers "where," dynamics (the next topic) answers "how fast can it get there given the forces involved" — they're complementary, not interchangeable.

Comparison and Connections

Concept	Similar To	Key Difference
Forward kinematics	Plugging numbers into a formula	Always has a unique, well-defined answer
Inverse kinematics	Solving an equation for an unknown	May have zero, one, or many solutions; requires checking workspace reachability
Kinematics	Dynamics	Kinematics ignores forces/mass; dynamics explicitly models them (see Robot Dynamics)
Revolute joint	Prismatic joint	Revolute produces rotational motion (angle variable); prismatic produces linear motion (displacement variable)
Workspace	Range of motion (biomechanics)	Same underlying idea — the reachable set — applied to a mechanical rather than biological system

Practice Questions

Recall

1. Define forward kinematics and inverse kinematics in one sentence each. Answer guidance: forward kinematics computes end-effector pose from known joint values; inverse kinematics computes the joint values needed to reach a desired end-effector pose.
2. What is a degree of freedom, and how many DOF does a typical arm need for full position and orientation control in 3D space? Answer guidance: one independent joint variable; 6 DOF (3 position + 3 orientation).

Understanding

3. Explain why inverse kinematics can have multiple valid solutions while forward kinematics always has exactly one. Answer guidance: forward kinematics is direct function evaluation; inverse kinematics involves inverting trigonometric relationships (e.g., cos(θ₂) = cos(−θ₂)), which can yield multiple joint configurations (elbow-up/elbow-down) reaching the same point.
4. Why can a target position be kinematically reachable but the arm still fail to reach it in the presence of obstacles? Answer guidance: workspace only accounts for joint limits and link lengths, not the actual physical environment; obstacles can block a geometrically reachable path even though the raw kinematic solution exists.

Application

5. For a 2-link planar arm with l₁ = 0.5 m and l₂ = 0.3 m, is the target point (0.9, 0) reachable? Show your reasoning. Answer guidance: max reach = l₁ + l₂ = 0.8 m; since 0.9 m > 0.8 m, the target lies outside the workspace and is unreachable.
6. A pick-and-place robot's vision system reports an object's (x, y, z) coordinates. Describe the sequence of kinematic calculations the robot controller must perform to actually pick it up. Answer guidance: use inverse kinematics to convert the target (x, y, z) [and desired gripper orientation] into joint angles, check the solution against joint limits/workspace, select a valid configuration avoiding self-collision, then command the joints to those angles.

Analysis

7. A robotic arm's inverse kinematics solver returns two valid solutions for a target point. What factors should the control system use to choose between them? Answer guidance: proximity to current joint configuration (minimize movement/time), avoidance of joint limits or singularities, avoidance of self-collision or collision with the environment, and smoothness of the resulting trajectory.
8. Compare the trade-offs of using 6 DOF versus 7 DOF (redundant) robot arms for a task in a cluttered workspace. Answer guidance: 6 DOF gives a unique or small set of IK solutions and simpler control, but limited ability to avoid obstacles for a given end-effector pose. 7 DOF (redundant) allows the arm to reconfigure itself around obstacles while keeping the end-effector fixed, at the cost of more complex, often numerically-solved IK and higher mechanical/control complexity.

FAQ

Q1: Do I need calculus to understand robot kinematics? Basic trigonometry and algebra are enough for forward/inverse kinematics of simple arms. Calculus becomes essential once you move to velocity kinematics (Jacobians) and dynamics.

Q2: Why do industrial robots almost always have 6 joints? Because 6 DOF is the minimum needed to place an end-effector at any arbitrary position and orientation in 3D space — 3 DOF for position (x, y, z) and 3 for orientation (roll, pitch, yaw).

Q3: What happens if a commanded target is outside the workspace? The inverse kinematics equations will have no real solution (e.g., you'd need to take the arccosine of a number outside [−1, 1]). A well-designed controller detects this and rejects or clips the command rather than sending garbage joint values.

Q4: Is inverse kinematics always solved with trigonometry like in the examples here? Only for simple arms with a "closed-form" solution. For arms with many joints or complex geometry, engineers use numerical/iterative methods (like the Jacobian-based Newton-Raphson approach) instead of exact algebra.

Q5: How is a singularity different from just reaching the edge of the workspace? They often coincide (a fully extended arm is both at its workspace boundary and in a singular configuration), but singularities can also occur inside the workspace for certain joint alignments, not just at the boundary — the defining feature is the loss of a degree of freedom of motion, not physical position alone.

Quick Revision

• Kinematics = geometry of motion, no forces involved (that's dynamics).
• Kinematic chain = links connected by joints; revolute joints use angle θ, prismatic joints use displacement d.
• DOF = number of independent joint variables; need 6 DOF for full 3D pose control.
• Forward kinematics: joint values → end-effector pose. Always unique, direct calculation.
• Inverse kinematics: desired pose → joint values. Can have zero, one, or multiple solutions.
• 2-link planar arm FK: x = l₁cos(θ₁) + l₂cos(θ₁+θ₂), y = l₁sin(θ₁) + l₂sin(θ₁+θ₂).
• Elbow-up/elbow-down ambiguity is the classic example of multiple IK solutions.
• Workspace = full set of reachable end-effector poses, shaped by DOF, link lengths, and joint limits.
• Singularity = configuration where the robot loses a DOF of achievable motion (e.g., fully extended arm).
• Denavit-Hartenberg parameters standardize link geometry description for complex, multi-joint arms.
• More DOF ≠ automatically better — it's a task-driven tradeoff between flexibility and complexity.

Related Topics

Prerequisites: Introduction to Robotics, basic trigonometry, vectors and coordinate frames.

Related Topics: Robot Dynamics, Control Systems for Robotics.

Next Topics: Robot Dynamics — once you know where the end-effector is and where it needs to go, the next question is what forces and torques are required to move it there.