Besides the machine learning angle discussed in the previous post, the UAV based GPR system that I have been working on has involved an interesting foray into robot swarming technology. The reason being that the sensor places a number of restrictions on the operation of the aerial robot. In particular the swath that can be covered in one pass is rather limited. There are different routes to ameliorating that and one of them is the use of multiple drones to increase coverage rates.
After meeting Sabine Hauert (of Robohub fame) at a Re.Work event that kernel of an idea evolved into a collaborative project with Bristol Robotics Lab and Skycap intern Pere Molins (who did all the hard work). The aim was build a demonstrator system that involved 3 aerial robots fly in formation and cover an area collaboratively in proper, fully decentralised, swarming style.
After some research we decided on the ErleBrain platform as a base and bought 3 ErleCopters to test with. The choice was not obvious as the brain had just launched and we were worried about stability. However the fact that it integrated ROS, Pixhawk, and BeagleBone in one single box made it very attractive and we decided to take the jump.
While we were waiting on the hardware, Pere, with help from Sabine, built a simulator to test the fundamental swarming algorithms.
This then progressed to software and hardware in the loop simulations followed by flight testing of the full hardware itself. Unfortunately it was then that we ran into trouble. The then current version of APM did not support the ultrasound sensors in the necessary flight mode and we had real trouble even maintaining a simple hover, let alone waypoint following.
This then led to custom APM patches, bug reports, forum questions, and much late night debugging. It was only just after the project had ended that the solution was finally found by…upgrading APM to new 3.4 beta release. It turned out there was some obscure bug we had run into in the flight controller itself. A pity, but such is life. The important thing is much was learned along the way.
Never the less Pere put together a great video that summarises the overall concept and vision:
Hence Queen B Robotics beat us with their demo video 🙂 but now that we have a base system Im curious to see if and how we can build on it further. The new Brian from Erle looks like a good step up and I think holds a lot of promise.