RoboticsROSMAVLinkPython

SUAS 2023

Built the planning and control software for an autonomous competition drone.

Role
Software and controls, end to end
Where
Project MANAS, Manipal's student robotics team
Year
2023

The result

2nd

SUAS 2023, Maryland

Planning and navigation software for an autonomous drone that placed 2nd at the Student Unmanned Aerial Systems (SUAS) Competition 2023.

How it started

Each year, Project MANAS builds an autonomous drone for the Student Unmanned Aerial Systems (SUAS) Competition.

For 2023, the competition made major changes to its mission requirements. The drone's software was written for the 2022 format and failed to adapt to the new standards.

The old system

The navigation planner was built under a tight deadline for the 2022 format, and it had multiple issues. It crashed often, and could not replan mid-flight or take new commands in the air. Its linear design meant every step had to be set before takeoff, in a fixed sequence.

A mechanically complex drone was held back by its software, which made progress on the platform close to impossible.

What got built

The planner was redesigned from scratch. The linear chain was swapped for a hub-and-spoke design, built around a central router.

In the old design, each step was its own component, but they were forced to run in a fixed line. Everything was uploaded to the drone once, as hard-coded instructions.

The router takes in the mission the same way, then calls each component on demand. New modules could be plugged straight into it. This made the design more flexible and easier to work on.

After: a central router connected to mission download, path planner, coverage planner, the drone, and two new modules: a release latch that drops the bottle and a position adjust that moves the drone left or right
Before: a linear chain of mission download, path planner, coverage planner and a one-time upload, ending in a drone with a hard-coded path
drag to compare the planner, before and after

Hitting the target

Due to a small expected error in an upstream component, the drone would always be a few meters off the drop location. That was not precise enough for a successful drop.

This is where the new design let us plug in a human-in-the-loop component. An operator tapped the target on the live video feed, and the component sent correction commands that lined the drone up perfectly over the drop location.

Result

After 100 to 200 test flights, the system was reliable enough to compete.

The team placed 2nd at SUAS 2023 in Maryland, USA.

I learned…

how to build a system that adapts when the task changes.