Autonomous Rover Obstacle Avoidance System

This project implements an autonomous obstacle avoidance system for a differential drive (skid-steer) rover using an OAK-D depth camera and MAVLink communication via mavlink-router. It uses depth data to detect obstacles in real-time and issues RC override commands to a Pixhawk (or similar) running ArduPilot.

---

System Overview

• Depth Sensing: Powered by the Luxonis OAK-D stereo camera using DepthAI SDK.
• Obstacle Detection: Custom-defined Regions of Interest (ROIs) in the camera view allow fine-tuned detection zones.
• Reactive Movement: Based on proximity and position of detected obstacles, the rover can stop, turn, or reverse.
• Control Interface: Sends RC override commands to the rover through MAVLink protocol via pymavlink.

---

Features

• Real-time obstacle detection using spatial depth data.
• Tunable obstacle response logic (STOP, TURN_LEFT, TURN_RIGHT, REVERSE).
• Safety-aware geometric check to prevent unsafe turns near obstacles.
• Configurable ROIs to match your rover’s field of view and size.
• Graceful exit on SIGINT/SIGTERM (Ctrl+C or shutdown).

---

Requirements

Hardware

• Pixhawk-compatible autopilot with ArduPilot Rover firmware
• OAK-D or OAK-D Lite stereo camera
• Companion computer (e.g., Raspberry Pi, Jetson Nano)

Software

• Python 3.7+
• depthai
• pymavlink
• mavlink-router

Install dependencies:

pip install depthai pymavlink

---

Configuration

Before running the script, tune the following parameters as needed:

MAVLink Settings

Parameter	Description
MAVLINK_ROUTER_HOST	IP address where mavlink-router is listening (usually 127.0.0.1)
MAVLINK_ROUTER_PORT	UDP port configured in mavlink-router (e.g., 14551)
MAVLINK_SYSTEM_ID / COMPONENT_ID	Target system/component IDs for your Pixhawk
GCS_SOURCE_SYSTEM	Source ID for this script (typically 255)

DepthAI Settings

Parameter	Description
DEPTH_RESOLUTION	Mono camera resolution (THE_400_P, THE_720_P, etc.)
DEPTH_RATE_HZ	Frame rate for depth processing
CONFIDENCE_THRESHOLD	Filter for noisy pixels (suggested: 220)

Obstacle Detection

Parameter	Description
OBSTACLE_THRESHOLD_M	Trigger distance for obstacle detection (e.g., 0.5m)
OBSTACLE_ACTION	Action taken on center obstacle (STOP, TURN_LEFT, etc.)

Motor Control

Parameter	Description
DRIVE_NEUTRAL_PWM	PWM value for motor stop (usually 1500)
DRIVE_FORWARD_PWM	PWM value for forward motion
DRIVE_REVERSE_PWM	PWM value for reverse motion
RC_CHAN_LEFT_THROTTLE_IDX / RIGHT_THROTTLE_IDX	Channel index for left/right motors

Rover Geometry

Parameter	Description
ROVER_WIDTH_M	Full width of rover in meters
SIDE_CLEARANCE_M	Buffer distance to prevent unsafe turns

---

How It Works

1. DepthAI pipeline is initialized with stereo cameras and spatial ROI calculations.
2. ROIs are configured to detect obstacles in the center, left, and right regions.
3. RC override commands are sent based on obstacle distance and location.
4. Geometric checks ensure turns won’t cause collisions with nearby obstacles.
5. The rover reacts accordingly by stopping, turning, or reversing.

---

Usage

1. Make sure mavlink-router is running and properly configured.
2. Connect the OAK-D camera to your companion computer.
3. Arm your rover and switch to Manual or Acro mode.
4. Run the script:

python3 obstacle_avoidance.py

Press Ctrl+C to stop safely at any time.

---

Testing & Tuning

• Use visual tools to debug ROIs and adjust their size and location.
• Start with low forward PWM and small ROIs.
• Increase SIDE_CLEARANCE_M if the rover gets too close to obstacles on turns.
• Test in simulation with SITL + MAVProxy if hardware is unavailable.

---

Credits

Built using:

• DepthAI
• ArduPilot
• pymavlink
• mavlink-router