Milestones from a road less travelled

Posted by: LightMetrics
Category: Uncategorized

For a technically complex end-to-end video telematics platform like RideView, stripping off as much of the complexity for end-users as possible, while still retaining the core essence of the underlying technologies should be a north star. For us, this has sometimes led us to take the harder, unconventional option, when easier (for us) options were readily available. Avoiding the easier choices, while difficult, has reaped outsized benefits in the long run. This is the story of some of those choices.

There are two distinct hierarchies of users that interact with our platform – Telematics Service Providers (TSP) and end-user fleets. Ease-of-use for us, therefore, has a wide definition, encompassing both ease of integration and final everyday use. In light of that, the following choices have helped reduce the time and effort that TSPs take to integrate our platform and enhance the experience of fleets in day-to-day usage.

ADAS auto-calibration:

Forward collision warning and lane drift warning are two of the major components of any ADAS stack, OEM or aftermarket. OEM solutions have one distinct advantage over aftermarket solutions though – the intrinsic parameters of the solution, from camera optics to exact height and orientation of camera mounting, are known in advance and fixed. These parameters are ultimately critical to calibrating a solution to providing accurate ADAS-based alerts.

Aftermarket video solutions do not readily provide such parameters to an ADAS system, unless installed by a professional. Professional installs though can be expensive, and cost-prohibitive for fleets with hundreds or thousands of vehicles. With the industry moving towards self-installable aftermarket dash cams, the other alternative is building UX for fleet managers or installers to mark critical key-points on the road-facing camera view, providing manual aid to the software calibration process. While cheaper than professional installations, doing this on even a medium-sized fleet of a few hundred vehicles can be time-consuming, tedious, and ultimately prone to errors.

For LightMetrics the problem is compounded by the fact that we support multiple dash cams, and TSPs or even fleets might be using different devices at the same time. Any manual intervention in the calibration process is, therefore, more complex, and error-prone. Realizing this, from the very beginning we have kept this calibration process completely automatic, and invisible to the end-user. Once a camera is installed, our ADAS software automatically detects salient features in a scene and computes the intrinsic parameters it needs to calibrate itself. Driving on roads with well-marked lanes under good visibility, at a speed greater than 30 mph, is the only pre-requirement, and this process usually completes in 5-10 minutes. The calibration results are stored and used for the device going forward. While this is technically a one-time process for a dashcam with fixed mounting, in aftermarket systems, there is always the chance that the camera orientation changes over time due to a variety of factors. For changes within a degree of tolerance, our auto-calibration routine re-computes the calibration parameters automatically on-the-fly, requiring no manual intervention.

Traffic-signs that are always up-to-date:

Speeding violations are one of the leading indicators of potential accidents down the line, and speed limit compliance is one of the key metrics that any safety system needs to track. Rolling STOPs, similarly, are a critical safety indicator, and our traffic sign compliance module covers both classes of signs.

It has been the norm in aftermarket telematics to license traffic sign data from mapping service providers to measure speed compliance. Such an option, though available to us as well, is one we have avoided in favor of detecting signs in real-time using AI, for the following reasons:

  • Third-party databases can be expensive, and more importantly, refresh at their own cadence. In real-life, signs can change,  signs can be removed, work-zone speed limits can come up and go, in a matter of weeks if not days. The refresh rate of standard databases is in months. This can lead to erroneous compliance reports, reducing trust in the system.
  • The ability for a dash cam to read traffic signs in real-time means that our data is always fresh, and drivers can be warned in advance to help drive compliance. Trust is also bolstered by the fact that, if needed, images and videos of speeding events can also be captured as proof, removing subjectivity in case of disputes.
  • As we have shared before, vision-based sign detection does have some fundamental limitations. If signs are occluded, or weather and lighting conditions make visibility poor, reliance on only real-time detection can limit performance. To compensate for these limitations, we have built our own crowdsourced database of traffic signs that we leverage to augment AI-based detection. This ensures that end-users get the best of both worlds – real-time and up-to-date data on speed limits, and robust performance come night or day, rain or shine.

Fully-featured fleet dashboards:

Our fully-featured fleet portal helps TSPs go-to-market quickly with white labeling. The easier approach for us would have been to provide the proverbial ‘picks and shovels in a gold rush’ (our APIs), and not get our hands dirty with end-user experience. While imagining the end-user experience is hard and warrants constant learning, feedback, and refinement, it is ultimately the correct choice. Our dashboards are an ideal showcase of the power of our APIs, the easiest way for our customers to launch, and a first step in iterating towards their own optimal UX.

Taking this a step further, to make navigating to our white-labeled instances seamless, we now help TSPs with both SSO and Iframe based integrations:

  • Single Sign-On (SSO): This allows a user to log in once and access various authorized websites or applications. Implemented correctly, SSO offers convenience and speed for users while ensuring adequate security and transparency. The authentication happens once (till logout – which can be through timeout or by user action). When a user logs into the parent TSP website for the first time, she gets prompted for a login-in prompt if there is no active session. On successful authentication against the central database maintained by the SSO provider, a token is sent back. This token is then used by the LightMetrics web application to authenticate the user and access services. The process of authentication is seamless and the user does not have to take any action. As long as the token is valid, or the user does not log out, there is no need to login each time one navigates to a new page.
  • Iframe: Using the Iframes feature in HTML one can display a webpage within another webpage. This is a common scenario for a TSP who wants to load multiple web pages managed by third-parties within its own website.  In the case of the LightMetrics fleet portal, it may be loaded within a TSP’s portal once a user logs into the parent website. In this scenario, the parent website returns a token on user authentication. This is handled through communication between the TSP website and the white-labeled LightMetrics portal that is going to load inside as an Iframe. This is implemented by listening to specific events and copying them into the local storage of the Iframe. Here onwards, the LightMetrics application running in the Iframe can keep using the token till it expires or the user logs out.

This list is not exhaustive, and there are a number of other choices that we have made along the way, motivated by end-user benefit, and without being daunted by the technical challenge. As our platform grows, and we support the needs of a diverse and global customer base, more such choices will keep coming our way. We intend to make the right one every time.

Header Image Source: Westside Church
Author: LightMetrics

Leave a Reply