IoT & Road Infrastructure To Fix the Autonomous Car Debacle

autonomous vehicles were all the hype in the 80s with night rider fast forward 40 years and voila we have avs but unlike kits in the famous series our versions still have work ahead of them as these cars have had a bit of a rocky start today there are over 1400 self -driving cars in the us that are being tested by over 80 companies so far in 2022 alone driverless cars have been involved in more than one 110 collisions, most of the accidents taking place in San Francisco, where robotaxis have already started operating on public roads. As autonomous cars are increasingly being released onto public roads across cities in the US, we turned to David Bruma, who worked on autonomous vehicle projects with the US He explained what issues autonomous cars are currently facing and the role existing IoT and road infrastructures can play to improve their operations. All of these so -called self -driving companies that tell and even use the word autonomy are not really autonomous if they require a connection to off -board maps, a connection to the cloud, a connection to GPS. And to be very clear, I am not saying that a connection to maps or a connection to the cloud or a connection to GPS is a bad thing. I just want us to be honest about the dependency. The real issue is. When I jam you, meaning I literally can go buy online a Chinese GPS jamming device, set it up for like $500, and what happens, you know what I mean? You lose your cell phone connection, use your GPS connection, can you function, can you not? When it comes to LiDAR-generated city maps, which are theoretically highly accurate and provide a solid base for the car's autonomous navigation, they are expensive to produce and not updated in real time and rely on cellular connectivity which means that on some occasions your map might be wrong and so whether it's wrong because there's snow banks all along the side after the show's been shoveled or because there's a change in construction patterns or because there's literally a forest fire that dependence on a map can actually end up going from being your reason for optimal behavior to your reason for disastrous behavior David told us about the issues related to off-board technologies so those inside the vehicle but what about the AI that gets fed by onboard technology such as radar LIDAR and other cameras don't they fill the gaps and figure out what to do without the help of maps or GPS I remember being a dark of the defense advanced research projects agency and get into some pretty hot debates over this topic because at that time the concept was and arguably still is that robots don't need any favors and that if we were just better ai guys that we would just figure it out robots should just figure it out why don't we just make them figure it out i mean come on um and my point was like well guys i mean humans are really smart but we put billions and billions and billions of dollars of infrastructure into our roads we put out street lights which is really hard we put all these marks on the ground we put signs all over the place like we didn't have that it would be a lot harder you know we don't just figure it out so stop saying that why are you saying that it sounds really appealing but it's not possible and it's also just dumb it's dumb we're not going to have an efficient road that way i know you want to do it i know you went to school to essentially write vision algorithms that require no assistance but it's not going to make for a good road it seems like there's a long road ahead of a b technology but how can we achieve real full autonomy? Is there a better way than what is being done now and how do we get there? We want a smart car we should sure as heck first have a smart road and a smart road what is that? Well it doesn't necessarily mean we have to make lots of expensive augmentations to the road I don't actually believe that we do necessarily. What I believe is that we need to have guarantees about position accuracy. we need to have guarantees about connectivity and we need to have essentially a way to embed command control and communication into the environment you know rf technology into the road on the road again i'm making it sound like this is really complicated like you have to bury stuff in the ground that's not true at all the irony is we already have streetlights why not take the photo cells for instance embed little chips into them so that we guarantee connectivity and we guarantee good positioning. Do you see what I mean? Absolutely possible today. Thousands and thousands and thousands of engineers who just frankly don't want to do it. They want to solve it with vision. They want to solve it with lasers. And this is deeply frustrating to me because I care more about the solution, which is lowering greenhouse gas emissions, improving the flow of traffic, making your commute shorter. That's the problem we need to solve. Not the problem of, hey, my big algorithm is really really good and i can use existing signs right so so what i envision is really what ash carter who's secretary of defense for the united states called the gps of things which is allowing these cars to really localize using each other as scientists um and to essentially understand their relative position peer-to-peer and also their relative position vis-a-vis the road and That is a very lightweight, very easy form of infrastructure because it really is, we're talking about, like, you know, tiny, tiny little chips. There are three key enablers to achieve full autonomy, according to David. One thing I think is essential is having relative position with no more than 10 centimeters of air. The second thing is that that ranging cannot and should not be limited to line of sight. If you have to wait for the car right in front of you. slow down before you slow down we've got problems the third thing is that we need to have communication with low latency in fact what we really need is we need to move towards single centimeters less than 10 centimeters and and microseconds and this is the point of the conversation where people say well that sounds very very ambitious no it's actually it's already solved like we have to do it, we're just not choosing to do it. Do you see what I mean? So we do need to invest in that interconnectivity and we need smart people who understand that need to architect it in the right way. As more autonomous vehicles hit the roads, we will keep investigating the technology and provide you with more information and opinions on the topic. In our next video on mobility, we will speak with the Virginia Tech Transportation Institute and UNIX Traffic. latest advances in v2x technology and we will visit pendor a barcelona -based startup to learn about their autonomous shuttle project in the port of barcelona so stay tuned on current insights