Up the Creek Without a Paddle

For many hobbyists, personal projects are not only about creating something useful, but are also about playing with new technologies and picking up some new skills. That is the way that embedded firmware engineer Braden Sunwold approached his recent DIY electric kayak project. Having spent a lot of time around large bodies of water for both professional and recreational reasons, Sunwold had long dreamed of building a custom electric kayak to give a paddling boost and extend his explorations. He has also been itching to get his hands dirty with edge machine learning, so he found a way to incorporate it into the electric kayak project, even though the combination of these technologies may not exactly be obvious.

Sunwold’s plan was to design and construct an electric kayak that can travel at about the speed of a person walking, and that has at least a two hour range so that it is useful for some good adventures. After investigating some computer vision models, Sunwold found them to be too challenging to implement in his project. So he instead came up with the idea to give his kayak an e-bike-like capability using machine learning. Many e-bikes will turn on their motor to give assistance when the rider begins to pedal. In a similar way, Sunwold wants the kayak to start its motor when the user begins to paddle. This is a more challenging problem in the water, because paddling is more than just propulsion — it also serves as the steering mechanism. That is where the machine learning comes in, along with sensor and GPS data.

This is a pretty large project, so Sunwold decided to split it up into phases, with the first phase involving the bare essentials — getting the electric kayak running, with a focus on speed and distance traveled. Sunwold started with an off-the-shelf inflatable Intex Explorer Kayak, then started collecting components to transform it. First, a waterproof, 500 watt DC brushless motor was selected. This is the same type of motor that is used in electric skateboards and hoverboards, so it should be able to provide plenty of thrust for the kayak. Being brushless, the motor also needed to be paired with an appropriate motor controller that could handle the motor’s maximum rated current of 12 amps.

Next, a fin, which also serves as the motor mount, was 3D printed and attached to the underside of the kayak. A pretty serious Lithium Iron Phosphate rechargeable battery was needed to meet the speed and distance requirements. Ultimately, a 24V, 20Ah battery was selected. Sunwold explains that this battery could power a typical iPhone for about six months straight. Connecting a battery like this directly to the motor controller would cause some serious sparking, which can cause problems over time. So, to address this issue, a custom anti-spark switch circuit was created. While commercial solutions do exist, the DIY approach saved quite a bit of cash.

Finally, to protect all of the electronics so that they do not get wet while out on the water, Sunwold put everything inside a waterproof case that looks a lot like the popular Pelican cases. After some bathtub experiments that confirmed the propulsion system was working and should be able to meet the speed and distance requirements, the hardware was fitted into the kayak. Sunwold then spent a day at the lake for a real-world test. Much to his surprise, the battery lasted almost four hours, or twice the range that was hoped for. This happy surprise resulted from some miscalculations related to the motor’s duty cycle, which was generally kept well below 100% in practice.

Phase one of the electric kayak project was a success. Stayed tuned for future updates where Sunwold plans to incorporate edge machine learning and more into the build.