Design-and-build project · confirmed

Design and build our own robot dog

A quadruped robot designed, built and tested by club members

Members will design a quadruped robot from a blank sheet, then build, assemble and test it themselves. Instead of buying a finished robot, the team works through every decision: leg geometry, actuators, the frame, batteries and wiring, motor control and the walking controller.

Building our own robot means members understand every part of it, can repair it themselves, and leave a documented platform that the next cohort can improve and use for guide-dog, delivery or inspection work.

Type
Design-and-build project
Status
Confirmed for 2026–27
Groups
All four technical groups
Goal
A robot dog that stands, walks and turns reliably, with full documentation

The robot, subsystem by subsystem

Each technical group owns one part of the robot; design reviews bring them together.

Legs and frame

Mechanical Design. Typically three joints per leg (hip abduction, hip and knee), a light and stiff body, and protected joints for falls.

Actuators and power

Electronics. Motors with low-ratio gearboxes and encoders, motor drivers on a CAN bus, the battery and an emergency stop.

State estimation

Perception. IMU and joint encoders to estimate body orientation and speed; a depth camera later for stairs and obstacles.

Walking control

Software & AI. A model-based gait controller first, then reinforcement-learning policies trained in simulation.

Engineering foundations

The equations behind a walking robot.

Leg inverse kinematics

\[\theta_2 = \arccos\!\left(\frac{x^2 + y^2 - l_1^2 - l_2^2}{2\,l_1 l_2}\right)\]

Knee angle needed to put the foot at (x, y) for thigh length l₁ and shank length l₂.

Joint torques from foot force

\[\boldsymbol{\tau} = J(\boldsymbol{\theta})^{\top}\,\mathbf{F}\]

The leg Jacobian maps the ground force at the foot to the torque each motor must give: the basis for actuator sizing.

Static stability

\[\mathbf{p}^{xy}_{\text{CoM}} \in \operatorname{conv}\{\mathbf{p}^{xy}_{\text{foot},i}\}\]

When walking slowly, the centre of mass must stay inside the polygon formed by the feet on the ground.

Gait duty factor

\[\beta = \frac{T_{\text{stance}}}{T_{\text{cycle}}}\]

The fraction of each step a foot is on the ground. Walking uses β > 0.5; trotting and running use less.

Project phases

  1. RequirementsSize, payload, speed and budget targets
  2. Design and simulationCAD, leg kinematics, gait tests in simulation
  3. BuildMachining, 3D printing, PCBs and wiring
  4. Assembly and bring-upJoint calibration, motor control, safety stop
  5. Walking testsStanding, walking, turning, then a public demo
Target: a robot dog that can climb stairs
Target: a robot dog that can climb stairs.
Gait training on rough terrain in simulation
Gait training on rough terrain in simulation.
Building and testing a robot platform
Building and testing a robot platform.