robotics

Where the pure feedforward story breaks: sensory entrainment, dynamic movement primitives for non-rhythmic motion, how CPGs combine with reinforcement learning in current legged robots, and the exoskeleton applications.

The experiment that put a spinal cord model in an amphibious machine: how one drive signal switches a robot from swimming to walking, and what that says about how vertebrate locomotion evolved.

How a handful of phase oscillators produce coordinated gaits: limit cycles and why they resist disturbance, phase coupling that locks oscillators into fixed relationships, and the travelling waves that swim a robot.

Walking is rhythmic, fast, and must survive disturbances no planner anticipated. The biological answer is a spinal circuit that generates rhythm without the brain, and it suggests a fundamentally different control architecture.

What happens when a prediction model leaves the benchmark: adversarial fragility, distribution shift, the feedback loop where your own robot changes the behaviour it predicts, and how planners consume uncertainty safely.

Displacement error is the field's default metric and it rewards the wrong behaviour: hedged average predictions, physically impossible trajectories, and best-of-many reporting that flatters diverse nonsense. What to measure instead.

How the field learned to represent people influencing each other: the physics-inspired force model, the Social LSTM pooling layer that replaced hand-designed rules with learned ones, and the attention and graph architectures that followed.

Why forecasting human motion is not a physics problem: the multimodality that makes a single correct answer impossible, the social conventions people navigate by, and the joint prediction problem where everyone is predicting everyone else.

How a vision-language model becomes a robot brain. This lesson covers the VLA recipe: take a pretrained VLM, add an action output, train on robot demonstrations. It walks through RT-2 (actions as tokens, web-knowledge transfer), OpenVLA (an open 7B model), and pi0 (flow-matching action experts), plus cross-embodiment, what generalization buys, and the real limits.

Two ideas that made imitation learning work: predicting a chunk of future actions instead of one step (ACT, from the ALOHA system) to blunt compounding error, and generating actions by denoising (Diffusion Policy) to capture the many valid ways to do a task. Plus the pooled multi-robot datasets, like Open X-Embodiment, that set up generalist policies.

Why modern robots learn skills from demonstrations instead of hand-written controllers. This lesson covers behavior cloning (supervised observation-to-action learning), the distribution-shift and compounding-error problem that makes it fragile, DAgger as the classic fix, and how teleoperated demonstrations became the fuel for robot learning.

How the Kalman filter fuses a motion model with noisy measurements by carrying a Gaussian belief, growing uncertainty on predict and shrinking it on update, weighting the two by the Kalman gain, plus the EKF and UKF for nonlinear systems.

An honest map of RL in 2026 — the domains where it actually reaches production (LLM post-training, robotics policies, ad bidding, RLHF, reasoning models) and the places where it still cannot reliably cross the lab-to-deployment gap.

Open-loop control is a guess; closed-loop control is a conversation. Learn why feedback transforms an unreliable robot into a reliable one, how the read-compute-actuate cycle works, what microcontrollers and single-board computers each do best, and why loop timing is just as critical as the algorithm running inside it.

A robot is only as good as what it can measure. Explore the full sensor toolkit — encoders, IMUs, ultrasonic rangefinders, LiDAR, cameras, and force/torque sensors — and learn the four metrics that determine whether a sensor is fit for purpose: resolution, range, noise, and sampling rate.

Strip any robot to its bones and you find three things: sensors that gather data, a controller that thinks, and actuators that move. Learn how these pieces fit together through the Sense-Plan-Act paradigm, why reactive control sometimes beats planning, and what makes a Roomba and a welding arm both qualify as robots.

Every robot motion — spinning a wheel, bending a joint, squeezing a gripper — starts with an actuator. Learn the key differences between DC, stepper, and servo motors; how gears trade speed for torque; how PWM lets a microcontroller dial motor power; and where hydraulics and pneumatics step in when electricity isn't enough.

When grids fail in high dimensions, random sampling saves you. Understand why PRM builds reusable roadmaps, how RRT grows a tree toward the goal, what probabilistic completeness really means, and how RRT* achieves asymptotic optimality.

Discretise C-space into a grid, then search it intelligently. Understand Dijkstra's optimality guarantee, how A* accelerates it with admissible heuristics ($f=g+h$), why consistency matters, and where greedy search goes wrong.

Understand why every motion planner secretly works in configuration space: how robots become points, obstacles inflate, and why high-dimensional C-spaces make naive search intractable.

Trace the full perceive-plan-act loop on a mobile robot: AMCL localization feeds a global costmap, a global planner (A*/NavFn) sets the course, and DWA or TEB local planners execute it — with recovery behaviors when things go wrong.

The chicken-and-egg problem of robot autonomy: to localise you need a map; to build a map you need a pose. Covers front-end matching, back-end pose-graph optimisation, loop closure, and visual vs LiDAR SLAM systems.

How LiDAR fires pulses and measures time-of-flight to build a 3D point cloud; data structures, voxel downsampling, ICP registration, ground segmentation, and an honest comparison with cameras for robot perception.

Fuse noisy sensors over time with provably optimal estimates. Covers the state and noise model, predict and update equations with the Kalman gain, why it's optimal for linear-Gaussian systems, and the Extended KF for nonlinear robots.

From photons to 3D geometry: the pinhole model, intrinsic matrix K, lens distortion, feature matching, stereo depth, and where CNNs help (and fail) in robot perception pipelines.

Learn how to generate smooth robot trajectories — trapezoidal velocity profiles, cubic and quintic polynomials — and how to combine feedforward and feedback to track them with minimal error on real arms and mobile robots.

Move beyond single-input PID to the state-space framework: the state vector, matrix dynamics, controllability, pole placement via state feedback, and LQR — the tool that scales to full robot arms and drones.

Master the proportional-integral-derivative controller: what each term fixes, the PID equation, integral windup, a discrete Python implementation, and a Ziegler-Nichols tuning guide for real loops.

Understand why feedback beats open-loop, how the classic closed-loop architecture works, and what the key performance metrics — rise time, overshoot, settling time, steady-state error — actually mean for a real system.