ROS Remote Surveillance Robot

An autonomous indoor surveillance robot built on ROS2, capable of real-time threat detection and autonomous navigation. Features a deep learning vision pipeline for anomaly and weapon detection, SLAM-based mapping, and remote monitoring via live camera feed.

---

Demo

📸 RViz + Gazebo screenshots, real world testing videos → https://tinyurl.com/Remote-Surveillance-Robot

Full Robot (RViz)	Gazebo Simulation	Physical Build

---

What I Built

This is a full-stack robotics project — hardware design, embedded firmware, ROS2 software, and AI perception all built from scratch:

• Custom PCB (Altium Designer) — Designed the robot's motor driver and power management board from scratch in Altium. Gerber files included. This replaced off-the-shelf boards and allowed direct integration with the onboard compute.

• Custom mechanical design (SolidWorks) — Designed the full chassis and sensor mount in SolidWorks. Supports two interchangeable wheel configurations: standard skid-steer wheels and mecanum wheels for omnidirectional movement.

• Dual drive configurations — The robot supports both a standard skid-steer wheel setup and a mecanum/omni wheel setup, each with its own kinematics node. This was a deliberate design choice to test different navigation behaviours.

• Skid-steer drive kinematics — Implemented a ROS2 differential drive node handling encoder-based odometry and motor velocity commands.

• SLAM-based autonomous navigation — Integrated SLAM Toolbox with Nav2 for real-time indoor mapping and autonomous path planning. The occupancy grid map below was generated from a live LiDAR scan on physical hardware.

• Deep learning vision pipeline — Integrated a YOLO-based object detection model for real-time anomaly and weapon detection from the onboard camera feed. Detection results are published as ROS2 topics for downstream alerting.

• Remote monitoring interface — Camera feed and robot telemetry streamed over the network for remote operation and live surveillance.

---

System Architecture

---

Hardware

Component	Details
Compute	Jetson & Desktop
LiDAR	RPLidar A1
Camera	Intel RealSense
IMU	MPU6050
Motor Driver	Custom PCB designed in Altium
Drive	Skid-steer / Mecanum (swappable)
Chassis	Custom SolidWorks design, 3D printed

---

SLAM in Action (Real Hardware)

The map below was generated live on the physical robot — not in simulation.

---

PCB Design (Altium Designer)

Custom motor driver and power management board. Gerber files available in /pcb.

---

Drive Configurations

The chassis supports two interchangeable wheel setups, each with a corresponding ROS2 kinematics node:

Standard Skid-Steer	Mecanum / Omni Wheels

---

CAD Design

View 1	View 2	View 3

---

Physical Build

Build 1	Build 2	Build 3

---

Getting Started

Prerequisites

• ROS2 Humble
• Nav2
• SLAM Toolbox
• OpenCV
• PyTorch / TensorFlow (for detection node)

Build

mkdir -p ~/surveillance_ws/src && cd ~/surveillance_ws/src
git clone https://github.com/The-Kriz/ROS_Remote_Surveillance_Robot
cd ..
colcon build --symlink-install
source install/setup.bash

Run

Simulation:

ros2 launch rog_ros_bot launch_sim.launch.py use_sim_time:=true

Navigation + SLAM:

ros2 launch rog_ros_bot navigation.launch.py

Surveillance / Detection:

ros2 launch rog_ros_bot surveillance.launch.py

---

Skills Demonstrated

ROS2 Nav2 SLAM Deep Learning YOLO OpenCV Altium Designer SolidWorks PCB Design Embedded C/C++ Skid-Steer Kinematics Mecanum Drive Python C++

---

Part of my robotics portfolio — see also Project EVA Humanoid Robot