By Alex Nesic
Over the last couple of years, micromobility has emerged as a preferred method of transpBortation in urban environments, naturally providing fresh air and physical distancing. However, that doesn’t mean there aren’t still kinks to be worked out. If you were to ask operators and cities the question, “What are the biggest issues with dockless LEV programs?” you would get some fairly divergent answers. These answers would touch on various themes such as operational efficiency, unit economics, regulatory compliance, data sharing, and insurance requirements.
‘AI-powered computer vision in particular has a valuable role to play in bringing together public and private interests in the evolution of micromobility as a long-term urban transit solution.’ – Alex NesicClick To Tweet
Behavior Management
Probably the most critical issue both sides would agree on is that user behavior management is an ineffective tool in producing the desired outcome in any scenario, be it following the rules of the road, wearing helmets, or parking (and sometimes locking) scooters where they should be. It is an undeniable truth that some humans find particular pleasure in eschewing (sometimes purposely) rules with which they are unfamiliar or deem inconvenient — in favor of greater personal convenience or expedience.
In the world of dockless micromobility, this produces less than optimal outcomes, sometimes involving injury or fatality, and often resulting in regulatory penalties and ADA violations. Ironically, the scrutiny under which operators and users of micromobility find themselves is draconian compared to the standards upheld for automotive user behavior under similar circumstances. I personally see this higher standard of safety and responsibility as an opportunity to solve some of these issues with technology.
AI & Computer Vision: The “Dash Cam” IoT Modules
The good news is that Artificial Intelligence (AI) is proving itself to be quite capable at mastering the type of boring yet impactful decision-making in the micromobility space that humans often seem unwilling to do — i.e. obey regulations. Adding to the human aspect of the problem is that these regulations often vary from one city to the next, making it even more difficult for users to comply consistently, even if they want to. Operators, data aggregators, and ancillary startups are already using AI to enhance their understanding of collected data with the goal of optimizing rebalancing, predicting demand, and performing preventive maintenance.
While these worthy pursuits will increase efficiency and profitability, I want to elaborate on a different type of applied AI — one that delivers edge-based ability for LEVs to recognize their position contextually in an urban environment, coupled with real-time vehicle control to optimize operational efficiency and prevent regulatory abuses. While this type of AI is ubiquitously used in the pursuit of Level 5 autonomous driving, it has not yet been applied to micromobility.
That is, until now. Companies are exploring the use of AI-powered computer vision as part of IoT solutions for shared micromobility. Traditional IoT modules that leverage GPS and cellular connectivity are already the norm in this industry, primarily used to locate, track and manage these free-floating vehicles.
Think of the AI and computer vision-enabled evolution as “dashcam” IoT modules that are capable of delivering granular infrastructure distinction in real-time. With the introduction of cameras, this new generation of IoT module can accurately and reliably distinguish between infrastructure categories like sidewalks, streets, and bike lanes. This feature alone dramatically enhances geofencing capabilities in a granular manner that existing GPS-based solutions simply cannot, particularly in dense urban environments where it matters most. The most pertinent application of this precise awareness is real-time sidewalk detection and the subsequent vehicle control it enables.
Edge-Based Solutions: Providing Critical Information
Adequate biking infrastructure falls short in most cities, and it is clear that additional investment in dedicated, safe bike lanes would produce better outcomes — as many cities did during COVID. However, even adequate infrastructure won’t stop scofflaw behavior entirely — bikes and scooters will still be ridden on sidewalks, and moped users will illegally use sidewalks, bike lanes, and park paths. Some argue that the incidence of sidewalk (or other illegal) riding is overblown, but the fact remains, solving for edge cases in transportation typically produces a more equitable overall solution and product — without detracting from the broader experience.
Beyond the regulatory and safety value of real-time vehicle awareness infinite environments, edge-based computer vision can be leveraged to verify proper parking in real-time where ‘corrals’ are used as designated parking areas. This technology reduces the need to rely on user-generated photos to end a ride and increases operational efficiencies. It can also be trained to recognize fallen scooters and flag them for corrective action — providing street teams immediate information to retrieve and relocate assets.
Algorithms can be trained to recognize additional, specific types of infrastructure based on operator and city needs — which can be leveraged to mitigate vehicle loss and theft. For example, a city DOT mused about the possibility of restricting scooter use in covered parking structures. Their concern was that scooters were being used to perpetrate theft and vandalize property in parked cars. For operators, this feature also has considerable value. When vehicles are ridden into such structures (or buildings), their GPS signal becomes significantly compromised, and this leads to many person-hours wasted in search of those vehicles — which are sometimes never located.
Contextual location awareness also provides valuable data insights to operators and cities. Currently, tracking of vehicle movements is only as good as the GPS information provided — computer vision can provide contextual distinction even beyond “lane-level” to include sidewalk, bike lane, and street-level route data. This data can be used by operators to better understand fleet behavior, by cities to help inform bike-friendly infrastructure planning, and by insurance companies to better assess risk and offer more dynamic pricing.
The Future of Micromobility
We now live in a drastically different world. COVID has accelerated the need for cities and operators to work together and align their objectives by collaborating more closely. This will allow for new climate-friendly modalities to flourish. Technology, and AI-powered computer vision, in particular, has a valuable role to play in bringing together public and private interests in the evolution of micromobility as a long-term urban transit solution.