Preparing For Easter Assembly

Børge Pahlm Project manager (Member since Sept. 2019)


Another week of progress!

This time there's more focus on the tasks parallell to the actual drones being built mechanically.

Beluga dipping in the tub for some testing by our Autonomous group (software).

The parts for the ASV (Autonomous Surface Vehicle) has been shipped for machining, and the first version of the new AUV (Autonomous Underwater Vehicle) has been machined. Now, there are a lot more parts to creating functional autonomous drones. Since last week we've bought a new desktop for CADing, tested the new pneumatic torpedo firing system, continued to work on the battery packs and sent all our chassis parts to machining.

Software lead Christopher Strøm working on the ASV software and simulator.

Software (but mostly perception)

The work put into the software development follows a much more continuous progress than hardware, or at least it seems like that from an observing point of view. Our technical lead for software, Christopher, has been putting in a great effort to supervise his software groups in addition to doing a lot of the work on the ASV codebase. The Embedded group is continuing to work on their challenges with localization using hydrophones while we're waiting for the new ones. The Autonomous group is currently testing Beluga in the pool, and the Perception group is working on new ways to filter the raw perception data from the stereo camera.

Highlighting the gate characteristics.

Pictured on the left is one the ways we're trying to extract the characteristics of the gate in the gate task of RoboSub using computer vision. The goal of this task has several steps beginning with recognizing where the gate is, approaching it and then go through it on the right side based on a predetermined route. Ivan, one of our Perception members has worked on a visual method. It involves equalizing the histogram of the original image from the stereo camera to boost the contrasts in the x and y axes independently, combining them and then using a blurring filter to reduce the noise. And voilà, a gate a appears!

Ivan (left) and Benjaminas (right) discussing the method.

This is one of the visual parts we're focusing on when using our stereo cameras. Next up is how we're looking into getting the depth data too! (Since there are two cameras like our own eyes!)

Stereo cameras are a thing that very recently got as commercially available as it is now, and we're planning to use them both on the AUV and ASV. It's an incredibly useful tool if you can get them to work right in relation to autonomy. They provide quality imagery, depth and other fancy sensor data from an IMU and other sensorics.


So, about that depth data.

A histogram of the camera image pixels (pointed at) and an associated frequency plot (left on the same screen).

The other part of the Perception group are trying to find ways to automatically detect when the stereo camera looks at objects in front of it based on its inherent detection of distance.

A sketch of how peaks in the plot implies existence of objects.

The video from the stereo camera is used to create a histogram of the pixels, and those are again plotted by frequency of depth. By evaluating the frequency of the placement of the pixels you can estimate the distance to an object in front of you. A spike in the frequency plot implies that there are is an elevated amount of pixels detected a certain length from the camera implying that there is an object in front of the camera.




Hardware

The Mechanical group is currently focusing on honing the mechanisms of our pneumatic actuator system involving a torpedo firing system, a release mechanism and a robotic arm all connected to the same air tank. As well as building their new PC. The Electronics group works on the new battery cells along with BMS (Battery Management System), the insides of the electronics housing on both drones and further improving PCBs (Printed Circuit Boards) for power and amplification.

To the right is our collection of differently designed torpedos that fits our conditions. We're using 3D-printed ones to quickly find the best mix between hydrodynamic design and amount of infill to glide through the water in the smoothest way possible.




When on the topic of 3D-printing, it is a greatly utilized tool at our workshop. Our printer works more hours than the average student I'd imagine. To the left is our current design of the insides of the electronics housing. Every year we complain about how difficult it is to maintain and repair our drones, and this is one step in the right direction. The structure will be placed permanently inside the waterproof housing, and all the electronics will be able to slide easily in and out on the rail in the red parts of it. All of the electronics will be attached to a board that fits perfectly between the two red parts in addition to being attached to the lid of the housing. This way we can unscrew it and just take out everything at once.



The battery packs are also an essential part of the work in electronics. We have used ready-made LiPo-batteries for a long time but have gotten tired of the restrictions they impose in regards to modifiability and safety. These new packs are made of Li-Ion batteries and hopefully will result in more stable, and standardized batteries, that we can use in more ways than before.


A BMS will be attached to each battery pack allowing for important diagnostics that the system will utilize during testing. Just this last summer one of our batteries stopped working during the TAC Challenge competition causing the drone to spin uncontrollably. At an ROV/AUV-competition like TAC where you're always connected to the drone it's incredibly valuable to always have the status of your batteries, and each specific battery cell, easily available.


And now that Easter is approaching fast (and the associated pool testing as well) the team's tempo is ramping up and we're looking forward to putting the new AUV together, and creating our first ASV prototype ever!