Modern Robotics Series Summary - From Geometry to Control and Planning

This series moves beyond viewing robot motion as a simple list of joint angles and instead treats motion as a geometric object. The central idea is that a robot moves through continuous rigid-body transformations, not merely through changes in joint values.

The full structure of Modern Robotics can be summarized as follows.

1. Starting Point: Why Coordinate Bookkeeping Is Not Enough

It is convenient to think of a robot's configuration simply as $q = [q_1, \ldots, q_n]^T$, but the end-effector pose is actually represented by a rigid-body transform

$$T \in SE(3)$$

where $SE(3)$ is the rigid-body transformation group containing both rotations and translations in 3D space.

This perspective matters because the end effector should not be understood only through changes in angles, but through the complete motion of a single rigid body.

The first step, then, is to view a robot not as a collection of assembled links, but as a sequence of continuous transformations of a rigid body moving through space.

Related reading:

• Modern Robotics, Part 1 — Rigid-Body Motion and the Product of Exponentials

2. Basic Building Blocks: Twists and Screw Axes

The next step is infinitesimal motion. A twist describes how a rigid body moves over an infinitesimally short interval of time.

$$\xi = \begin{bmatrix} \omega \\\ v \end{bmatrix}$$

Here, $\omega$ is the angular velocity and $v$ is the linear velocity. A twist is therefore a compact representation of the instantaneous velocity of a rigid-body motion.

This twist can also be expressed through a screw axis that combines the rotational and translational components.

$$\mathcal{S} = \begin{bmatrix} \omega \\\n-\omega \times q \end{bmatrix}$$

This matters because each joint does not merely change an angle; it generates rigid-body motion along a specific axis in space.

In other words, a joint is a motion generator, not just a scalar variable.

Related reading:

• Modern Robotics, Part 2A — Screw Axes, Twists, and the Exponential Map
• Modern Robotics, Part 2B — Frames, Lie Groups, and the Jacobian Bridge

3. Finite Motion: The Exponential Map

We also need to describe finite motion, beyond infinitesimal changes. This is where the exponential map becomes useful.

$$T(\theta) = e^{\hat{\mathcal{S}}\theta}$$

This equation is central to many ideas in robotics.

• Given an infinitesimal motion generator $\hat{\mathcal{S}}$,
• and a joint value $\theta$,
• the final transform $T$ is obtained by exponentiation.

This is the core idea behind the Product of Exponentials (POE). The robot's complete pose is constructed by composing the screw motions of its joints.

The robot's pose is therefore understood not merely as an operation on coordinate frames, but as a composition of joint motion generators.

4. The Jacobian: From Joint Rates to Task-Space Velocity

The next key connection is the Jacobian. Given the joint velocity $\dot{\theta}$, the end-effector twist is

$$\nu = J(\theta)\dot{\theta}$$

where $J(\theta)$ is the Jacobian.

The important point is that the Jacobian is not just a matrix; it is a geometric mapping between joint velocities and end-effector velocity.

The columns of the Jacobian are the screw axes generated by the individual joints. The Jacobian therefore collects all of the robot's geometric motion directions.

This perspective lets us understand:

• velocity control
• manipulability
• singularity
• velocity tracking

Related reading:

• Modern Robotics, Part 3 — Jacobians and End-Effector Velocity
• Modern Robotics, Part 4 — Manipulability and Velocity Control

5. Inverse Kinematics: Finding Joint Values for a Target Pose

To make a robot reach a desired pose, we must solve the inverse kinematics problem.

$$T_{se}(\theta) = T_d$$

This problem is generally nonlinear and may require either an analytic or a numerical solution.

A common approach is a Jacobian-based iterative method.

$$J(\theta)\Delta\theta = \nu_e$$

The joint values are updated repeatedly. This process uses local linearization to bring the robot's pose progressively closer to the target.

For a redundant manipulator, we can use the pseudoinverse.

$$\dot{\theta} = J^{\dagger}\nu_d$$

The nullspace can also be used to achieve a secondary objective:

$$\dot{\theta} = J^{\dagger}\nu_d + (I - J^{\dagger}J)\dot{\theta}_0$$

This allows the robot to maintain the primary task while improving its internal configuration.

Related reading:

• Modern Robotics, Part 5 — Inverse Kinematics and Numerical Solvers

6. Dynamics and Trajectory Generation: Real Robots Do More Than Match a Pose

Reaching a target position is not enough. A real robot must move smoothly while satisfying acceleration limits, inertia, gravity, collision avoidance, and timing constraints.

The dynamics can be written in the form

$$M(\theta)\ddot{\theta} + C(\theta,\dot{\theta})\dot{\theta} + g(\theta) = \tau$$

.

The Jacobian also describes the relationship between task-space force and joint torque.

$$\tau = J(\theta)^T F$$

It connects forces in task space to the robot's joint torques.

Trajectory generation is not simply a matter of specifying target poses; it is the design of a smooth motion profile. A trajectory must satisfy pose, velocity, and acceleration limits, while motion planning finds a collision-free path through the environment.

Related reading:

• Modern Robotics, Part 6A — Dynamics, Control, and Motion Generation
• Modern Robotics, Part 6B — Trajectories, Constraints, and Motion Planning

7. The Complete Flow in a Single Set of Equations

The central structure of this series can be summarized as follows.

$$T_{se}(\theta) \in SE(3)$$ $$\xi \in se(3)$$ $$T = e^{\hat{\mathcal{S}}\theta}$$ $$\nu = J(\theta)\dot{\theta}$$ $$T_{se}(\theta) = T_d$$ $$M(\theta)\ddot{\theta} + C(\theta,\dot{\theta})\dot{\theta} + g(\theta) = \tau$$

These equations connect the major themes of robotics:

• geometry: rigid-body motion
• differential geometry: twists and screw axes
• dynamics: inertia and torque
• control: Jacobian-based velocity control
• planning: feasible trajectories and collision avoidance

8. Conclusion: The Central Message of Modern Robotics

The most important idea in Modern Robotics is not a collection of formulas, but a structured way of understanding robot motion.

Rather than being a simple engine, a robot is a system built by composing continuous rigid-body motions. Those motions can be distilled into the following concepts:

• rigid transform
• twist
• screw axis
• exponential map
• Jacobian
• inverse kinematics
• dynamics
• motion planning

Once this complete flow is understood, robotics becomes more than implementing motion in code; it becomes the study of designing motion through geometric structure.

This series was both a first step into that subject and a summary of its larger picture.

The next step is to connect this mathematical structure to real robot software and control code. Many details remain, but they can all be understood on top of this foundational structure.