Design Philosophy
The main aim of the project is to have a novel way to practice and learn about machine learning. The key to this is therefore collecting as much data as possible. The boat records all data every second and stores it locally, keeping log files, and periodic data is reported to mission control! This data will be essential to see how well the boat performs, to enable new features to be designed, to detect faults, and to implement machine learning.
Physical Design
The Hull
Design
Although catamarans are more stable, if they do capsize they are much harder to right, therefore a monohull design has been chosen. To create an efficient design I wanted a streamlined and small hull. Imagine pulling a large box through the water compared with a surfboard. The surfboard displaces less water, there is less resistance, and it is easier to pull along. The width of the boat has to be large enough to accommodate solar panels, but taking inspiration from aircraft carrier design, the bottom of the hull is narrow for efficiency, with a larger top deck. A narrow hull isn’t very stable, so towards the stern the hull gets wider and flatter in order to provide stability.
Construction
A 3D model was designed on the computer and then 3D printed to check for errors. Using this model as a template a plug was created, then a mold, and from this the hull. A fiberglass construction was used due to its strength and light weight properties. 4 layers of fiberglass were used (300gsm chopped strand mat, 600gsm woven roving and 300/600 biaxial combination mat).
Performance
The boats LWL (length of boat at the waterline) is 2.5m which is important as a boats length determines it’s maximum speed. A displacement hull would have a top speed of 3.8kts and a semi-displacement hull would have a top speed of 7.1kts. Therefore I’ve tried to design a semi-displacement style hull as high top speeds may be required to avoid other boats and beat currents.
The Keel
A 15Kg keel 0.8m from the bottom of the hull has been used. Due to the top heavy nature of the design, this is essential to keep the boat stable and upright. The keel bulb is currently torpedo shaped however a wing shape may be used to increase efficiency.
Motor and Rudders
If it’s to make it to the arctic and back it needs to be resilient. So where possible critical components will have a secondary in case of failure. Although, this is a balance between cost and redundancy. The boat will have dual independent motors and rudders. If the rudders fail, the motors will be able to apply differential power to steer. If one motor breaks down, the other motor can be used with extra rudder to compensate and keep it on a straight course.
The motors will be placed in pods underneath the boat to ensure they are always in contact with the water. As the boat is relatively light it doesn’t sit far down in the water, so if they were mounted out of the stern, the traditional way, the concern was that they wouldn’t always be in the water!
As for propeller choice, read more here to learn about choosing the right motor and propeller combination.
Mast
A mast sits in the centre which serves two purposes. The primary purpose is to house 2 cameras and antennas at the top of the mast. However to supplement power from the solar panels a vertical axis wind turbine will also be used.
Power
Power is stored in 6 x 5Ah lipo (lithium-polymer) batteries, providing 24Wh of power. Power is sourced from 2 x 100w solar panels and a vertical axis wind turbine (VAWT) will provide additional power. Although the VAWT will produce drag and pull the boat off course it is anticipated that the extra power gained will outweigh this disadvantage.
Whilst the motors will use most of the power, a lot of power will be required for other hardware such as the Raspberry Pi. The Pi does all the controlling including image processing (which is computationally intensive) and AIS signal processing.
If you connect a solar panel to batteries, it will not operate at it’s most efficient point. Therefore an MPPT device will be used as this finds the solar panels optimum operating point, and provides the most power to the batteries.
Hardware and Software
Arduino
The arduinos main purpose is to monitor and control power. The voltage and current of the solar panels and batteries are measured at all times, and the arduino will turn the charging or main power off or on accordingly. The arduino can also be used to read analog sensors which the raspberry pi cannot.
Rasperry PI
The Raspberry Pi is the brains of the boat, that will make decisions and control the boat. The boats systems have been programmed in Python and the only part it is not responsible for is power management.
Navigation
To navigate, the boat initially follows a series of waypoints. As it follows the route it will calculate drift produced by currents and the wind and attempt to compensate for this. Many sensors are used to help the boat navigate. GPS tells the boat where it is and an IMU (Inertial measurement unit) can tell direction, orientation and acceleration. This is useful as the direction will tell us if the boat is facing forwards, but if the GPS direction tell us the boat is moving backwards, we know the current is too strong! As the IMU shows orientation and acceleration we can see how the boat moves in the water. This can be used to calculate wave height, so we can see if it’s in a storm, or measure vibrations on the boat so if something starts to fail, hopefully we can detect the problem and get it back to land quick!
To avoid other boats, and eventually icebergs, it will use a combination of AIS and cameras. AIS is a transponder commonly found on large boats which emits information such as vessel type, location, speed and direction. We’ll modify a TV tuner stick to receive this information. This information will become vital as we can combine it with the camera data to teach the system what type of boat it is looking at, so when a boat doesn’t have AIS, the system can do it’s best to predict what it is seeing. When a boat is detected on AIS it will work out if we are on a collision course and make the appropriate corrections.
There will be 2 raspberry pi cameras, which will be used for stereoscopic imaging. This will allow the boat to calculate how far away different objects are. If the cameras sense a hazard a course will be created to steer around it.
Communication
There are several different methods of communication used. When the boat is nearby 434Mhz radio can be used for low latency communication. This will be useful to control the boat manually when testing in a lake as careful control is required for launching and landing. For greater distances, if it’s nearshore 3G can be used. This allows the boat to send large data such as image files and other data for analysis. The last method used is satellite, as there is almost global coverage, however it’s expensive, slow, and can only handle 250 bytes maximum. That’s not a lot of data, to put it in perspective the italics paragraph above is all that can be sent, and would cost £0.75. So if we want to report it’s location and other stats every few hours, we’re going to have to get clever with the data.
When data is sent from the boat it needs to go somewhere to be stored! AWS is the answer, that’s Amazon Web Services. Data is sent to an API which then gets decompressed, checked and processed by a Lambda (serverless code). This then writes the data to a DynamoDB and to flat files in S3, ready for analysis.
There are some safety features built in, so when controlling the boat manually via radio, if it goes out of range, the engines will cut out. Also if a ‘hello’ is not received from mission control over a certain period of time the boat will change course and sail home. This is so if something goes wrong with the communications or computing, the boat returns home automatically and doesn’t float randomly for the next 10 years!
Motors
2 x 300w brushless motors are used due to their improved efficiencies over brushed motors and they are powered by 2 x ESC (electronic speed controllers).