Drover AI’s Alex Nessic, Co-Founder and Chief Enterprise Officer, joins Ryan Chacon on the podcast to debate the function of AI and pc imaginative and prescient inside IoT. Alex begins by introducing himself and his firm earlier than speaking concerning the limitations of GPS. He then describes how AI and pc imaginative and prescient works collectively to resolve the constraints. The dialog then turns extra high-level with dialogue round the way forward for pc imaginative and prescient and the potential of shifting into the patron area.

Alex Nesic is a passionate evangelist for shared micromobility and LEVs usually, selling their function in a sustainable city transportation ecosystem. Alex is co-founder of Drover AI, pioneering the usage of AI-powered pc imaginative and prescient for micromobility. Drover is his third firm within the micromobility area – Alex beforehand based CLEVR Mobility and was an govt at Immotor.

Concerned with connecting with Alex? Attain out on Linkedin!

About Drover AI

Drover AI is pioneering the usage of AI-powered pc imaginative and prescient on IoT units used within the micromobility trade. Their PathPilot tech is utilized by our prospects to reinforce their regulatory compliance whereas additionally serving to them optimize operational effectivity. PathPilot makes use of a digicam and Drover’s AI algorithm to detect in real-time whether or not a scooter is touring on a avenue, sidewalk, or bike lane, enabling management of the automobile’s pace in every space. On the finish of rides, PathPilot additionally performs parking validation, serving to enhance the right parking end result cities want and protecting the right-of-way unencumbered.

Key Questions and Matters from this Episode:

(01:21) Introduction to Alex and Drover AI

(05:11) Limitations of GPS

(10:02) How AI and Pc imaginative and prescient work collectively

(13:52) Way forward for pc imaginative and prescient and IoT

(16:40) Shifting into the patron area

(18:41) Roadblocks within the trade


– [Voice Over] You’re listening to the IoT For All Media Community.

– [Ryan] Howdy everybody, and welcome to a different episode of the IoT For All Podcast, the primary publication and useful resource for the Web of Issues. I’m your host, Ryan Chacon. In case you’re watching this on YouTube, please be at liberty to love this video and subscribe to our channel. In addition to in case you’re listening to this on a podcast listing, please be at liberty to subscribe to get the most recent episodes as quickly as they’re out. All proper, on right this moment’s episode, we have now Alex Nesic, the Co-Founding father of Drover AI. They’re an organization that’s pioneering the usage of AI-powered pc imaginative and prescient on IoT units used within the micromobility trade. We discuss AI and pc imaginative and prescient and their function in IoT. How does it work, how does all this enhance security throughout the board, what issues does it actually clear up, the way forward for pc imaginative and prescient and its function in IoT, in addition to different challenges type of seen throughout totally different areas because it pertains to micromobility and IoT, AI, pc imaginative and prescient, all that type of great things. So plenty of worth right here. I believe you’ll get pleasure from this episode a ton. However earlier than we get into it, if any of you on the market want to enter the fast-growing and worthwhile IoT market however don’t know the place to start out, try our sponsor, Leverege. Leverege’s IoT options improvement platform, offers the whole lot you might want to create turnkey IoT merchandise you could white label and resell underneath your individual model. To be taught extra, go to iotchangeseverything.com, that’s iotchangeseverything.com. And with out additional ado, please get pleasure from this episode of the IoT For All Podcast. Welcome Alex to the IoT For All Podcast. Thanks for being right here this week.

– [Alex] Thanks for having me.

– [Ryan] Yeah, it’s nice to have you ever. Let’s go forward and kick this off by having you give a fast introduction about your self to our viewers.

– [Alex] Positive, my title is Alex Nesic. I’m a Co-Founder and Chief Enterprise Officer at an organization known as Drover AI. And we’re pioneers within the area of bringing pc imaginative and prescient, so cameras and edge-based synthetic intelligence machine studying to IoT units, particularly within the shared micromobility area. So assume shared e-scooters and bikes, issues like that.

– [Ryan] Unbelievable. So yeah, why don’t you… Let’s elaborate somewhat bit extra on what the corporate does. So take us by way of perhaps a use case software of the expertise in a setting that’s relatable for the viewers.

– [Alex] Yeah, completely. So something that’s deployed as a free floating asset, the best way that these scooters and bikes are out within the subject, requires an IoT module. And the first function for that’s to trace and handle them. And the connectivity depends on GPS sign and a chip onboard, in addition to mobile connection, making it a related gadget, proper? And so forth the again finish, the operator can monitor and handle the place the entire units are situated, battery standing, upkeep standing, something like that, quite a lot of different… And from a buyer’s level, that’s the way you act with the gadget by way of the app, proper? That’s the way you find the scooter and the bike and lease it, , pay, all of that. It’s all managed by way of the IoT module. What Drover AI brings that that’s actually modern is a further layer of intelligence. At the moment GPS, , has plenty of margin for error, particularly in, in dense city environments the place plenty of these fleets are, are operated the place tall buildings will distort the GPS. You already know, and so your location functionality is proscribed by what, , the, the, the accuracy of the GPS, which is, , purposely a reasonably type of moderately priced off the shelf module. And so as a substitute of, of simply counting on GPS, what our system does is brings a digicam to bear and edge based mostly computing energy to run our algorithms. And we offer what we name contextual location consciousness, that means that fairly than counting on a exact GPS based mostly location, we place ourselves extra like a human who goes exterior. And doesn’t say, what are my coordinates? You go searching and you’ll see that the road is perhaps 10 toes away, that you simply’re on the sidewalk, that there’s a motorcycle lane there. So it’s within the context of a location. And so even when we occur to be in a, or the automobile occurs to be in a spot that’s devoid of GPS connectivity, or mobile connectivity, the truth that it’s all occurring on the sting permits our IoT module known as the Path Pilot to establish the place it’s in, within the context of its surrounding. So whether or not it’s on a avenue, sidewalk, or bike lane through which permits the operator to, to do quite a few issues. First, amongst them being regulatory compliance, it’s unlawful to be on sidewalks with scooters. So our expertise allows the management of the scooters pace or, , conduct and even surfacing audible notifications to the person to allow them to know that they’re in violation of, of a regulation.

– [Ryan] That’s cool.

– [Alex] We additionally assist with, with parking, proper? And these free floating fleets, the tip of trip is an enormous concern as a result of customers typically depart them in the midst of the suitable of method, which is in violation of the Individuals with Disabilities Act. And so it, it simply helps our, our prospects, the operators, handle these units way more safely and responsibly and, and it helps fulfill the calls for of the town.

– [Ryan] Unbelievable. So one factor I wanna truly develop on, I, I do wanna dive into extra of the pc imaginative and prescient and the function it performs in IoT, however you talked about one thing about GPS. We speak lots about GPS and totally different type of applied sciences on the subject of monitoring belongings and, and issues like that. However inform us, or, or give us somewhat bit extra element as to what, what downside, I assume, I assume what GPS doesn’t essentially clear up for itself and type of your capacity to type of slot in along with your expertise and your providing in these densely populated areas in these cities. And inform us somewhat bit extra concerning the limitations of GPS in that setting and type of what you’ve been capable of do to, to optimize and make it type of believable now.

– [Alex] Yeah, completely. So GPS, once more, , off the shelf chip set offers you some accuracy wherever between three toes, all the best way as much as 30 toes. And the difficulty there’s that it doesn’t degrade gracefully when it’s challenged. So in case you open up your cellphone and also you take a look at your little blue dot, google maps. If you end up in a, a metropolis with actually tall buildings, you’ll discover that that blue dot has type of like this aura round it, or a halo that’s giant, that signifies it’s fairly unsure of its location. And then you definately type of see it bouncing round till it will get a increasingly and extra exact, , thought of the place it’s truly situated. However even in that state of affairs, you might be coping with not simply sign attenuation, however multi-path, which is mainly when a sign is bouncing off of a constructing, plenty of buildings are glass and stuff. So that you, you, your precise exact location or blue dot won’t even be reflective of the place you truly are. It’s decoding the place the sign is bouncing off of and placing you perhaps throughout the road. So there’s all of those totally different challenges that GPS, , has a bunch of methods up its sleeve, enhanced GPS methods, which require earthbound base stations known as RTK, realtime kinematics, which basically is a distributed community of, of earthbound stations that might enlarge or amplify a GPS sign for, for enhanced place. You may use mobile phone triangulation, you may use useless reckoning and different sensors on board to attempt to mainly sew collectively the trajectory of, of one thing that’s shifting and useless reckoning does that. However all of that type of finally falls quick in probably the most difficult areas. And, and what’s necessary there’s, is that in case you’re attempting to tell apart, , GPS works properly for these broad areas that you might want to geofence. Perhaps you desire a geofence a whole metropolis block or a, a college campus that doesn’t need scooters there. So GPS is nice as a result of it doesn’t matter in case you’re 20 toes off, proper? I imply, if, in case you cross the boundary and, and the, the deactivation of the scooter solely occurs 20 toes later, it’s not a difficulty. But it surely is a matter in case you’re attempting to tell apart between when a person has left the road and entered the sidewalk, that’s, you want like, , a number of, a pair inches accuracy. There to essentially play in that type of finite area. And so what pc imaginative and prescient actually does there’s convey that, that digicam and skill to contextually place itself and never depend on, on a GPS sign that is perhaps fairly far off for all intensive functions. After which the opposite a part of GPS, proper, that, that doesn’t scale essentially quickly is that allow’s say you do obtain that type of holy grail of accuracy of 10 centimeters or higher in all environments, which is uncertain, however let’s just- Hypothetically say you could, what then the problem turns into like a extremely huge information administration problem, since you nonetheless don’t have the context of the place you might be. You’ll have pinned down your correct GPS location, however you continue to don’t know the place that places you in, , the place the road ends and the sidewalk begins. And so what must occur is it’s known as a floor fact layer. You must exit and gather that information and draw an entire bunch of strains on some kind of, of backend Google maps or different to establish, , what that infrastructure is at these actual GPS coordinates. And so no metropolis has that details about itself. You must exit and gather it and create that database. You must replace it over time as cities evolve and sidewalks would possibly fluctuate- Or different infrastructure fluctuates. And so it doesn’t scale simply. It’s a, it’s an enormous information administration nightmare throughout a number of areas and what pc imaginative and prescient and, and edge based mostly AI onboard the IoT gadget can do is, , even when our gadget has dropped into an entirely new atmosphere with out having any floor fact, as a result of it positions itself contextually, it doesn’t want that info. It could possibly accomplish that by, , leveraging a basic database of hundreds of thousands of photos throughout dozens of cities that will look moderately just like that. And so we will infer based mostly on, off of that current database the place it finds itself and make that call in actual time.

– [Ryan] Gotcha. Unbelievable. I’d like it If we may transfer out somewhat larger degree right here for a second. Only for our viewers to get a way of, after we’re speaking about pc imaginative and prescient and the AI and the way it all type of works collectively. Inform us somewhat bit extra about even exterior of your basic use instances And focus space, how that, how pc imaginative and prescient and AI type of collectively are enjoying a task in IoT. How does that work? How is that enhancing security type of, as we’re speaking about right here concerning the totally different ranges of accuracy it permits us to type of get to, however simply, simply break that down somewhat bit additional for us so we will type of simply perceive it a bit higher.

– [Alex] Yeah. I believe it’s actually, , innovative expertise right here in, in these type of creation of, of pc imaginative and prescient based mostly networks which are related, proper? I imply, you possibly can see a really type of related software is the RingCentral community, for instance.

– And, and these are mainly static sensors which are leveraging these cameras on doorbells or different safety techniques which are deployed, and so they create this mesh community basically the place you, you possibly can faucet into them and thru movement detection or different type of clever sensors and sensor fusion onboard. These have a extremely correct perception into what is perhaps occurring in our neighborhood. You already know, definitely site visitors administration and different deployed sensors exist already on, on lamp posts and, and cease lights in, in cities. Properly, I believe what’s distinctive about, , what we’re doing is, is deploying these on type of cell belongings that transfer by way of cities very otherwise, clearly otherwise than any static sensor, however otherwise than automobiles do that may already be mapping a sure atmosphere. And so cameras actually convey that, that further layer of context that with out it, you, you wouldn’t actually essentially have that, that understanding or, or capacity to, to take motion. So an instance could be, , Google maps leverages information from their autos shifting by way of cities. However the common age of that info is round 18 to 24 months outdated. And so how do you plug in, , perhaps more moderen info and, and so we will perhaps leverage our distributed belongings shifting by way of cities, simply on client rides to say, Hey, at this timestamp and placement, we have now an information set. You would possibly be capable of, to faucet into that’s more moderen than what, what has been collected 18 months in the past. So there’s, there’s a ton of, of actually neat, related digicam networks which are being constructed out and, and that may very well be leveraged for various makes use of.

– [Ryan] Yeah. That’s tremendous fascinating to type of take into consideration simply how we will make the most of these distributed belongings to gather info to simply profit plenty of totally different functions from an information perspective, proper? Like with like Waze, as an example, you drive round in Waze and when Waze first began, they might actually simply use the, the, the GPS location of the automobiles that had the early Waze app and say, okay, in the event that they’re driving right here, there should be a highway right here. After which finally map software program and stuff bought built-in in and issues like that. However having the ability to use these scooters and different sorts of distributed belongings to gather info that we’re not capable of gather as precisely is tremendous fascinating.

– [Alex] Yeah. Properly, and truly to, to develop on that. Your, the, the feed is delayed on the video. To develop on that, you possibly can layer in a bunch of various issues like object detection, proper? Let’s say you wish to do some infrastructure surveying. What are the circumstances of my sidewalks? You already know, what’s the foot site visitors like right here, clearly taking privateness into consideration right here the place we will masks and blur people, however, however yeah, there’s an entire number of several types of machine studying datasets that can be utilized on the info that we’re gathering to, to assemble further insights.

– [Ryan] Completely. Yeah. It, it it’s tremendous fascinating stuff, which I assume it expands into a extremely good type of comply with up query is what does the way forward for this sort of appear like? Like let’s develop on the place it’s now and, and the chances to type of the way forward for pc imaginative and prescient, its function in IoT, and type of the place you see it going exterior of simply these use case we’ve been speaking about.

– [Alex] Yeah. And, , our core competency and, and our, our preliminary prospects are the shared micro mobility. And, and it’s particularly addressing mobility wants and regulatory compliance and security. Total, our aim is to essentially, , improve the adoption of micro mobility for a extra sustainable city transportation future that strikes away from automobiles. And, and in order that’s actually the place any additional pursuits of, of monetization, of, of information that we’re gathering would ideally be used to subsidize the expansion of, of micromobility, both by way of further infrastructure that may be constructed out. Bike lanes and networks that may accommodate extra of some of these, of autos. However yeah, you alluded to, we’ve already touched on a few of them. The place the digicam and, and, and IoT related units can drive insights associated to city trans or transportation planning, city planning, . We will add different sensors on board, air high quality sensors for instance, that, that might be comparatively low-cost to, to include into these distributed belongings so that you simply’re gathering air high quality samples from totally different non-static sensors, shifting by way of a metropolis. That’s all the time been fascinating. However in case you, in case you additionally wish to go and, and perceive the place sure building is going on, that may not be mirrored. On, on some other metropolis dashboard, proper? We will begin recognizing the place, , cones are deployed and, and different site visitors deviation patterns are, are occurring to tell, , mapping, proper? And, and, and move by way of, by way of a metropolis. We will inform curb administration selections, proper? If, if we will establish, , that, {that a} curb area is getting used improperly, or, , once more, this entails some public, non-public partnership, proper? I imply, however we will additionally do the identical factor that, that bus fast transit lanes are, are doing the place they’re leveraging cameras onboard buses. To routinely ticket violators, individuals which are, which are utilizing the bus fast transit lane after they’re not alleged to, we will do this for bike lanes, proper? I imply, so if, as a metropolis you’re allocating assets to construct out a motorcycle community, however persons are utilizing the bike lane as a supply zone. That’s not efficient. It’s not protected. It pushes individuals out into the road. We may leverage our automobile detection in addition to some type of license plate recognition and, and be capable of ship citations or on the very least warnings to those that repeatedly do this.

– [Ryan] Do you assume it’s ever one thing you could transfer into extra of the patron area so far as automobiles clearly have cameras on them, whether or not it’s sprint cams. Or simply automobiles, cameras inside the autos, is there ever a stage through which that may be leveraged from a, from, for the info type of within the general profit, whether or not it’s security or simply type of basic coordination of issues. And I, I simply don’t know if there’s essentially like perhaps some particular parameters on why that’s not perhaps both authorized or allowed, and that’s why it’s extra public kind alternatives for this information to be collected simply outta curiosity.

– [Alex] Yeah. No, I believe you convey up level. I believe we’re, we’re all nonetheless evaluating alternatives on this area. You already know, an organization that involves thoughts while you discuss that, which mainly has adopted what, what you’ve described as a enterprise mannequin is, is Nexar the place they leverage client sprint cams that, which are deployed in client autos. They usually subsidize the price of these sprint cams by gaining access to, or retaining entry to the info, to have the ability to create these dashboards and, , perceive various things and supply insights to cities or different prospects. So I believe there’s, there’s all the time been this fascination with the right way to monetize or what the worth is of, of information that comes off of autos, every kind, proper? There’s lots of of sensors on each client automobile that’s put on the market and corporations like Autonomo, and Wejo are, , presently valued at, at very, , excessive valuations based mostly on the premise that this information that, that they’re ingesting partially, , by way of partnerships with OEMs has worth and, and, and totally different insights could be derived from them. So I believe we’re nonetheless very early days when it comes to what precisely could be accomplished with several types of information. Each simply sensor information, in addition to, as imaginative and prescient based mostly information. So we’re, we’re actually enthusiastic about totally different alternatives.

– [Ryan] Completely. Yeah. It’s, it looks as if the chances are fairly countless, which is basically cool to see. Final query earlier than we wrap up right here is simply out of your perspective of the market and type of the way you guys are viewing the IoT area, what are a number of the different huge challenges you’re seeing corporations face, perhaps corporations that you simply’re talking to immediately, corporations that you simply’re simply type of, type of maintaining a tally of and simply usually talking, operating into potential roadblocks with IoT adoption, deployments, issues like that.

– [Alex] Yeah, certain. So one of many largest ones now that I believe plenty of corporations within the electronics sector are coping with their provide chain. And logistics, proper? Availability of elements. There’s been a pressure on, on that attributable to the pandemic and, and the ripple impact that that’s had. So I believe definitely with the appearance of IoT, totally different IoT initiatives which have grow to be wildly widespread. I imply, one other one which involves thoughts is, is Helium, which is a, , a LoRaWAN, lengthy vary WiFi community successfully of hotspots that, that has, , put a, a robust demand on, on type of compute modules which are utilized in, in IoT units. So, so I believe, , usually provide chain and logistics is a, is an enormous one. One other one particular to pc imaginative and prescient is basically navigating the privateness points, ensuring that you’re as an organization and, and as a expertise CCPA and GDPR compliant with, , redaction of photos on the edge earlier than they’re ever shared with anyone. So we, these are the type of obstacles and challenges that we have now to think about as we attempt to scale our enterprise. Yeah.

– [Ryan] Completely. Yeah. It’s some tremendous fascinating, simply type of, while you go from provide chain to chip shortages, simply the whole lot occurring in, within the area, the way it’s affecting IoT, I imply, it’s affecting tons of various industries, however simply, simply seeing how, how that’s occurring. It’s been a extremely fascinating dialog with totally different visitors have had on the podcast simply to grasp how they’re coping with it and what they’re seeing from, from their very own perspective since every space of IoT is feeling it, however perhaps feeling it in somewhat bit totally different method. For certain. Final thing earlier than allow you to go right here is for viewers on the market who needs to be taught extra about what it’s that you simply’re doing, the expertise use instances type of, perhaps it applies to issues that they’re engaged on, and so they’d like to type of develop additional and contact base. What’s the easiest way to try this, to comply with up and type of keep in contact.

– [Alex] Positive. Our web site is www drover.ai. It’s place to be taught somewhat bit about what we do. I’m additionally on LinkedIn, Alex Nesic. You’ll be able to comply with me on Twitter @Alexnesic and the corporate can also be on Twitter @droverAI.

– [Ryan] Unbelievable. Something new, thrilling type of popping out of Drover that we ought to be on look out for within the coming months. And simply, simply something you can provide us a sneak peek on.

– [Alex] Yeah. We’re working with increasingly new prospects. We’re deploying in bigger scale in Europe.

– Cool, cool

– Largely with Voy, certainly one of our largest prospects there. So we’ll be spending a while supporting the launch of a number of lots of and hundreds of, of path pilots on the market.

– [Ryan] That’s superior

– [Alex] After which we’re additionally, , type of shifting to our subsequent era product, which is gonna be built-in immediately into autos, not bought as a retrofit add-on module. In order that’s type of the subsequent step in our, in our transition is to insert ourselves additional up within the provide chain and already be type of a baked in choice in, within the automobile itself.

– [Ryan] Now let me ask what’s the worth and profit there, versus having the ability to type of do it as an add on. Is it simply the deeper type of relationship with the corporate and the product, or is there clearly function and, and information advantages from that?

– [Alex] No, it’s primarily prices, proper? I imply, as a result of proper now. Our out on module is, has some redundancy with what already exists. Each certainly one of these autos already has an IoT module that has GPS and mobile connection. And so we additionally need to have that on our IoT module. So now you could have two units of GPS charges, two units of mobile modems, et cetera. And so no person needs to be paying twice for a similar factor. So I believe having the ability to, to combine and at, on the edge with these sensors that exist already and successfully simply leverage these onboard techniques can be a a lot less expensive option to introduce this. However, , a secondary module was, was actually a, a choice that we took to, to have the ability to go to market as rapidly as doable and show out the expertise at scale ‘ absolutely figuring out that we might finally be integrating into any individual else’s automobile design. However in case you, in case you begin off on that path, you’re a minimum of 18 months away from any kind of market deployment. So our, our alternative was to say, Hey, right here’s a tool you possibly can slap on any scooter within the subject and we’ll work with you on that. And we bought to market lots sooner by. By selecting that path.

– [Ryan] Makes complete sense. Unbelievable. Properly, properly, Alex, this has been an awesome dialog. Thanks a lot for taking the time, very excited to type of keep, keep in contact and, and see all of the various things you could have popping out and kinda the evolution of Pc Imaginative and prescient and, and the way issues are evolving in your finish. So I’d like to have you ever again. I’d like to kinda get you concerned in a few of our different sequence as properly, that type of align with what you could have occurring, however thanks once more a lot in your time. Actually respect it. I believe our viewers is gonna get a ton of worth outta this.

– [Alex] Superior, Ryan, thanks a lot for having me. It was a pleasure talking with you and I stay up for any future conversations we’d have.

– [Ryan] Sounds nice. All proper. Take care. All proper, everybody. Thanks once more for watching that episode of the IoT for All Podcast. In case you loved the episode, please click on the thumbs up button, subscribe to our channel and you’ll want to hit the bell notifications so that you get the most recent episodes as quickly as they grow to be accessible. Aside from that, thanks once more for watching and we’ll see you subsequent time.


Subscribe Us to receive our latest news in your inbox!

We don’t spam! Read our privacy policy for more info.


Please enter your comment!
Please enter your name here