I know I've fallen silent over the last few weeks, with rushed blogs posted now and then, and honestly, that's because I keep forgetting to post them. I don't forget because I don't care, but rather that my head is filled with nonsense to do with this project. This blog is dedicated to all that nonsense, and hopefully, by writing it all down it will become less nonsensical, and maybe, just maybe, I'll be able to understand just what the hell I've let myself in for. To quickly summarise previous blogs, my idea for my University honours project is a blind-spot detection for motorbikes. Not in the sense of where the kit detects when there is something in the bikers blind spot, but when they are sitting in a cars blind spot. Bikers and cyclists are 65x more likely to be or seriously injured/killed in a crash than drivers and more often than not caused by other road users that forget to look out for people on bikes. My idea is designed to be more of a training aid than a rider aid, helping to promote safe riding and teaching 'the bubble technique' from advanced rider training, a technique that encourages beings fully aware of all 360* of your surroundings, keeping an appropriate distance from vehicles and judging what different drivers could do if they haven't seen you. A wee while ago I posted a survey into a few Facebook groups, and within a day, it reached over 370 responses. Out of those respondents, roughly 67% had been in a collision. I asked people for details of their crash, and only one person wrote about the 'bubble technique' and how because of this they'd never been in a collision. Now, I'm not claiming that my idea is going to prevent all crashes on motorcycles, I can't even come close to making that statement as the physical product doesn't exist yet. Still, I do believe that by training motorcyclists to be more aware of their surroundings and creating good habits within beginner riders, will, in turn, result in a decrease of motorcyclist accidents. So where am I now? Just last week I had a meeting with a senior lecturer in Computing, Science and Engineering to see if he knew a way I would be able to get my device to recognise blind spots, and thank god he did. I knew that it technically would be possible, Tesla and Volvo use similar technology to detect when vehicles sit in their blind spots. Hence, all I'd need to do is take this and kind of flip it around to work in the opposite direction. But I'm not Elon Musk; I'm Jamie, a 21 uni student that doesn't even really understand his idea. Thankfully, the lecturer managed to break down how to get my idea to work, and it is possible at my level. I was already somewhat aware that I'd need to use something called Artificial Intelligence, but having never worked with it before I wasn't sure if it would be something I could learn in only a few months. This lecturer broke it down and made it easy to explain. He called it 'Machine learning' and said I'd need to use a software that would be able to recognise, after learning what one is, what a blind spot looks like. Take cats and dogs, for example. If I wanted to teach a machine to recognise a cat or a dog I'd first need to show it what they looked like and I'd do that through a series of pictures. I'd have to upload photos of different cats and dogs of different breeds, shapes, sizes, colours etc. to a machine learning software and specify what each one was. Eventually, this software would learn how to recognise the differences between a cat and a dog, so when an image of a cat or dog is shown to the machine, it will be able to recognise what species of animal it is to a reasonable accuracy. For the technology to become more accurate, more examples need to be uploaded to the software. Take Tesla's self-driving cars for instance (slightly different technique but the basics are still there): when the self-driving vehicles first hit the road, all they had to base judgement off what the information uploaded for Tesla's tests. Over time, as more people bought the cars and tried out the autopilot feature, data was collected from over thousands of different experiences, learnt by the software and shared with all the vehicles, allowing each car to become more 'intelligent'. There are open-source, free software that I can use to replicate this software, and I won't have enough time to make it hugely accurate. If I really wanted to make it work, I'd have to buy a RasberryPi and extra ram for it, as Arduino's aren't powerful enough. I also want to create a way for riders to be review all the information collected while out riding with my product, so right now, and the idea is to email feedback after each journey. The email will include states like the percentage of time spent in blind spots, what kind of blind spots they spent time in (trucks and lorries differ from cars etc.) and how they can improve their riding and avoid these areas. The idea is, as more people buy or use the product, the system will learn more about road safety, blind spots, and be able to provide more accurate feedback on each journey and how to improve in different situations. Going forward, I'm not going to focus 100% of my time on getting the technology to work. I want to make sure that the concept is solid so firstly I need to focus on the experience of the product. How will the system warn the rider? Will it be a flashing light or sound? Will it be on the bike or the helmet? How will the email be laid out? These are things that will all affect the experience of the product before the technology will, and by working on all of this (even if the technology doesn't work) I'll still have something to exhibit at the Degree Show. Well, that's been a long one today. It's currently 18:56 as I write this, I'm still in the studio, and I'm starving, so I'm calling time after this. Moving forward, I hope to experiment with different ways to feed information to the rider so that I will be focusing on a lot of prototypes between now and the next major milestone. Thanks for reading and I'll see you all next time!
Cover Photo Image: Pidrsushnyi, D. (2019) Person riding on motorcycle. Available at: https://unsplash.com/photos/tbE-Y16gGtg (Accessed: 02/03/2020).
Comments