Sailing Auto Pilot - Competition - Ready For Tomorrow 2022

ok who won the prises?

I've already worked on ROVs and UUVs. Since this is not an under water vehicle, I've some suggestions that can be added to the existing design. So the first thing is to integrate an SBC like Nvidia Jetson Nano or Raspberry Pi. This will add good computation power to the robot. Then coming to the software architecture, my suggestion is to deploy ROS on SBC. Since I'm a ROS Engineer curently working on Autonomous UUVs, I've seen in real time that ROS is robust for making this kind of system into an intelligent machine in a cost effective manner. And next thing is sensor fusion, since the robot moves on the surface of water, we don't need to wory about buoyancy control things. So we can easily focus on sensor fusion part. For that we need an IMU sensor and GPS module and using filters like EKF we can fuse the sensor data to get the accurate pose of the drone. When coming to the obstacle detection part, we can go with 2 sensors. One is Lidars, to detect the obstacles like another boat near by the robot or some other obstacle over the surface of sea. And another one is sonars, which can be mounted under the robot to detect obsacles that is below the surface of water. So Inorder to make the system to avoid obstacles we are going to utilize the Lidar and sonar sensor and control the thruster and rudder of the bot. As far as I understand, It seems like arduino is used for controlling the actuators. So we can utilize the ROSserial communication protocol to make the SBC to talk to the Arduino. So we need to create a script by utilizing the lidar and sonar data and then make the actuators move accordingly based on the obstacle detection.Now coming to the autonomous part, we can use GPS waypoint based navigation by integrating all the features that I've mentioned previously. By this way, the system will be more robust and at the same time will be intelligent. And my final suggestion is creating a custom dashboard, with speedometer dial, battery level indicator, compass and orientaion dials. This makes the user to understand about the bot's current heading direction and all other essential informations. Hope my suggestions helps to develop the robot into a better and intelligent one.

A thing about the display. You should be able to read this in full sunlight. On plotters for navigation they use 500+nits color screens. Those are expensive and power hungry. You could use a waterproof e-reader like the kobo h2o. It is water proof and has a good screen, backlight and wifi. It saves you from having to build a waterproof case. You can probably get one cheap second hand.  

I think while suggesting many advanced auto-pilot system details, we all forgot the main thing in an auto-pilot system... the manual override switch. Just in case the control has to be transferred to the user, if the autopilot system fails or otherwise. 

Another suggestion to improve the autopilot. I would add a strain sensor on the mainsail as shown in picture

The autopilot will read the force "exerted" by the mainsail. A drop or change in the detected strain means that the sail is flapping due the the wind flow starting to detach from the sail surface. This can trigger some corrections from autopilot to get back to the optimal course. When the sailor sets an heading on the autopilot, the autopilot itself will read the current force detected by the strain sensor and will continuously try to keep this level of force by slightly correcting the course. A proper algorithm needs to be implemented to prevent an excessive drift from the desired heading  

Here are some examples of how the autopilot can leverage the feedback provided by the strain sensor

1. prevent the mainsail from crossing the boat in case of sudden changes in wind direction. This safety function could be automatically disabled by some switches on the helm that detect that the maneuver has been intentionally initiated by the sailor
   
2. prevent damages due to strong wind: when detected forces raise above a safety threshold, the autopilot could correct the course to go closer to the eye of the wind, thus reducing the stress on the sails
3. when sailing closed-hauled, correct the course to make the boat stay on the "sweetspot" on the border of the "no-go zone". More precisely, when the sail starts flapping, the strain sensor will detect a drop in force exerted by the sail and will correct the course to go automatically closer to the wind

Very nice project !
It's funny, we are working on a somewhat similar project with a sailor who has done the Mini-Transat 6.50 (Edwin on the mini 6.50 Nacira "Coeur Fidele", 7th production boat in 2015)
We are developing an autopilot mainly based on the raspberry Pi4, open source, and maybe we could suggest some ideas...
Some info here (not really up-to-date):
https://sailing.netcys.com/
The autopilot must support different algorithms, from the simple PID to more advanced solutions : we are testing solutions using "auto learning" from monitored information from sensors (wind, heel, speed, ...) and computing power of Pi4 is really required ...

Another idea for Alberto's project:
something very useful for solo sailing, inexpensive, and easy to build, is a voice alarm reporting system that monitors NMEA status. Alberto's boat is equipped with NKE electronics, and the easiest way is to use the Topline-NMEA output option.
The “voice synthesis alarm reporting system” assists the skipper to monitor the main parameters available on the boat, from the different sensors: wind, depth, loch, …
Alarm messages are played on the Audio output to be mixed with the audio output of the on-board VHF, and to be transmitted to the cockpit speaker.
For example, a significant change in wind angle, or wind speed, etc. generates a detailed voice alarm, which allows the skipper to take better advantage of short periods of solo sailing rest ...
We had made this simple piece of equipment with a Raspberry Pi3, and with a very inexpensive little color screen for the Raspberry for the Mini Transat 2015 ...
When solo sailing, voice synthesis alarm may be more useful than a simple sound alarm ...

Because the autopilot is not allowed to change the sail trim, it could change the direction to optimize the sail trim.

Thanks to a digital telltale, the autopilot will be able to know the efficiency of the sail trim in relation to the direction.
If the wind turns, and the direction is more favorable, then the autopilot will change the helm in that direction.

Technically, a digital telltale would be composed of 2 telltales (a piece of fabric), one on each side of the sail.
Each telltale is composed of a Bluetooth Low Energer (BLE) transmitter and a gyroscope at its end. The autopilot has basically to compare the tilt of the 2 telltales (thanks to the data of the gyroscopes transmitted in BLE) in order to know if the sail trim is optimized and if not it will also know in which direction to go to optimize it.

Passionate about sailing and embedded development, I will develop this project in open source with great pleasure!

Good suggestions for the auto pilot system however, GPS is not allowed as per the rules

Hi rsjawale24. Thanks for your comment. Since Alberto was going to user a smartphone to remotely control the autopilot, I thought to exploit the same wireless communication channel (I proposed BLE, but wifi is fine too) to send route adjustment commands. I don't think this is a safety-critical application: this is just a "gadget" to make sailing more efficient by letting the sailor react faster to changes in wind direction. If a gesture is not detected or there are delays in communication, the boat will simply maintain the current route and the worst that can happen is that you have to reach the autopilot and change the course using the local commands. On the contrary, a wireless man overboard device is a safety-critical device and, as, such, should be properly engineered and validated. Finally, I proposed gestures because a gesture requires just one hand. This means, for example, that you can change course while operating a winch

Also, in a system like this it is nice to keep things modular. So you have the ability to replace parts. Things break or get wet and burn out. Once on my autopilot the drive transistors did burn out on my Raymarine autohelm. Luckily I had an old one laying around and could desolder the transistors of that one. It was no race so I had time. But things break when you push them to the limit, and then having a spare is nice.

As in system stability it could be a good idea to run this on a dedicated mcu. Then you have complete control, development time will increase thou. Systems like this should be very stable. 

Also in this race electronic map software is forbidden. If you run a system that could use an electronic map, like OpenCPN but does not because it is against the rules. The jury could rule against allowing the system because you could install it easy even when you are underway. If the mcu does not run linux, openCPN would be impossible. 

Linking the system to a vhf is not allowed but there is a transponder to keep track of the ships. You maybe could hook into this system in case of an emergency. Later when the race is finished you can upgrade the system to include the vhf and ais.

As for the practical side. It is a small boat so I guess the lineair drive is not so very power hungry. (Blowing out the transistors of the Autohelm was on a 22ft sailing boat, so about the same size). To keep things modular and linux free the arduino shields could work: the A000079 motor driver with 4A drive capacity. A GPS like the MIKROE-1714 and a IMU like the MIKROE-1577 and 10212635-00 for weather sensing. Your boat has a wind vane and wind speed sensor, those work with NMEA2000 if it is a modern sensor, for this you will need a CAN shield:MIKROE-4107.

As for the MCU go or something like the H7 lots of space speed and connectivity. 

As for the rest of the electronics you can put it on a base board where you can connect all the modules together. If something breaks just replace the module.