SCOUT Quadruped Robot

SCOUT (Search and Communication Operations Using a Tetrapod) began as an affordable robotic dog for learning how mechanical design, robotics software and AI work together, and is now a proof of concept for search-and-rescue support. I'm building it with two teammates for a RASTIC project at Boston University.

After a disaster, debris, stairs and narrow openings make it hard for responders to reach people who are injured. SCOUT is a proof of concept for how a small, low-cost robot could help: a 12-DOF quadruped with a compact 4-DOF arm that can explore a simulated disaster area, help identify a person who may need assistance, report back to an operator, and reach into spaces the body can't fit.

The project started as something simpler, an affordable robotic dog for learning how mechanical design, robotics software and AI fit together, with a parts budget of $500. I'm building it with two teammates: one builds the dog's body and legs, and one leads the AI and software.

I own the 4-DOF camera arm. That covers the arm and gripper, its electronics and wiring, a compact folding arrangement (folded onto the top of the robot or into the body while it walks), the arm's access to the supply compartment, and mounting the visible and thermal imaging system with the software lead. I'm also exploring a foldable solar-panel mechanism with my mechanical teammate. Right now I'm designing the camera mount in CAD.

The robot has four 3-DOF legs for walking patterns, body-height changes and uneven surfaces. The arm carries a camera and a gripper to inspect narrow spaces and retrieve small supplies. Thermal imaging sits alongside visible imaging because a heat signature alone doesn't confirm that a person is present.

SCOUT is also meant to talk. A microphone and speaker let an operator speak with a person near the robot, over WiFi for demos, with onboard processing and a local link as the longer-term goal. A simple operator GUI lets the operator start a task, view images, receive alerts, check the battery and connection, and stop control.

A supply compartment, holding items such as bandages, sits under a camera that tracks which item the arm removes. Foldable solar panels are being explored as a future option for supplemental charging, to be weighed against their added weight, size and mechanism complexity.

The chassis is a cylindrical concept inspired by the enclosed layout of a submarine, and it is being compared against other layouts for stability, assembly, cooling and repair access.

The planned hardware is a Raspberry Pi 5 (16 GB) as the main onboard computer, and a PCA9685 servo driver for the 16 joints (12 in the legs, 4 in the arm) with a separate servo power supply. An NVMe SSD holds the OS, ROS and sensor data, and an IMU provides orientation and balance. An INMP441 microphone and a MAX98357A amplifier with a speaker handle audio. An Arducam IMX219 USB camera sits on the arm, a TF-Luna LiDAR checks arm clearance, and a NEO-6M GPS is there for future outdoor tests. The control software is planned around an inverse-kinematics gait, ROS and simulation.

The build order is deliberate: test one leg and the arm on their own first (motion, servo loading, wiring and control), then assemble the chassis and four legs, then integrate the arm, cameras, audio and supply compartment, then build the operator interface, and finally run controlled demonstrations.