Road Condition Monitoring and Remote Comms

This page contains links to AWS solutions, road condition monitoring and edge computing / comms links

Data Sources

The US Department of Transportation provided open access to a wide range of relevant data and methodologies for traffic, road and transportation analysis and management

Here’s a report titled Integrated Modeling for Road Condition Prediction [Final Report] https://rosap.ntl.bts.gov/view/dot/35421

  • There are statistical norms for road usage
  • There are spectral density curves for speed and traffic density that classify roads
  • Not shown in this report but power spectral density (PSD) curves of vehicle vertical acceleration (G) against frequency also provide useful classification of road condition (base excitation) on vehicle motion
  • The tech used in this study is quite antiquated but the data sources and statistical analysis are key takeaways
  • Lessons learnt are key takeaways in preparing to tackle the problem of road condition monitoring at scale
  • Pavement state prediction can be obtained from satellite imagery, vehicle imagery and secondary methods like variations in traffic flow and variation in speed from PSD diagrams

Open Data

There are some sample road data sets available but it’s challenging to find large scale data sets that include road condition data.

Vehicle Data Acqusition

The options include:

  • Instrumented vehicles (expensive and fragile)
  • OEM sensors (canbus)
  • Vision based (mix of on vehicle camera and retrofit)

Another option that works surprisingly well is leverage Smartphone data from accelerometers, gps, barometer, temp sensor

AWS Solutions and Partners

This section describes key technologies to support road condition monitoring at scale

Spatial Platform

There are relevant learnings, and solutions, from spatial humanitarian work that is relevant to road condition monitoring at scale.

The Orbital Summit presentation at the AWS Public Sector summit in Washington DC in 2022 titled Geospatial intelligence: Supporting military analysts & humanitarian aid https://www.youtube.com/watch?v=Z0dpo28uNfs

  • Dual data sources are needed. Spatial and on vehicle sensors combined allow road condition to be determined from vehicle (micro) or from above (macr0 - space, drone, aircraft imagery)
  • AI is crucial for imagery to determine patterns and insights
  • location data from vehicles allows areas of interest to be identified
  • The general geospatial analytics platform describes the data, sensors, data processing, analysis and insights on a page.
  • The publically available metadata about people (in this general geospatial analytics platforms) could be replaced with restricted access unit or personnel information. There is a ‘security piece’ here that is not depicted.

Digital Twins

The general geospatial analytics platform could be the precursor to a broader digital twin that has a broader business IP benefit. Data is the ’new oil’…

Blog post titled Digital Twins on AWS: Understanding “state” with L2 Informative Digital Twins https://aws.amazon.com/blogs/iot/l2-informative-digital-twins/

  • Starting at an L2 Informative Digital Twin as a POC, would leverage the general geospatial analytics platforms to demonstrate value and prove out data sets and AI, ML and Analytics methods for the customer
  • As the solution matures then it can be expanded into an L3 - Predictive and ultimately an L4 Living Digital Twin to deliver shareholder value

Smart Manufacturing / Production & Asset Optimisation

Connecting many devices and collecting data at scale is supported by numerous AWS Solutions. The Machine to Cloud Connectivity Framework is one example https://aws.amazon.com/solutions/industrial/

  • This solution can be extended to support remote comms and edge technologies broader than what is in the base solution
  • This solution scales and has baked in security and is cost effective at scale uising serverless technologies.

AI and ML Road Condition Assessment

Satcom and Edge Processing

Present thinking in the EV space is that vehicles will transmit 7+TBytes of data per day. This thinking does not address issues in low bandwidth connectivity and on vehicle processing.

Things get complicated at scale:

  • Satcom connectivity and data transfer is expensive. Although Starlink is / has addressed this issue. Others are trying to emulate.
  • Edge Data and Edge Comms devices will invariably be in numerous states of obsolescence, needing updates, unreachable, etc. You need to be able to manage this type of edge fleet. AWS IoT services address this from the software defined level but the tracking and management of hardware is a traditional asset management problem. Connectivity with edge dependency (like physical SIM) is problematic, although LoraWAN is a more elegant solution. Defence use cases require ‘quiet vehicle / no RF transmission’ modes which discounts much commodity hardware.
  • Low power LoraWAN edge devices often lack sampling rates needed for road condition monitoring for vehicle mounted sensors. (1 - 5Hz for sprung mass and 8 - 15Hz for unsprung mass)
    • which means you need 10 - 100 Hz sampling rates to determine road condition from unsprung mounted sensors
    • 1 - 10 Hz sampling rates will resolve sprung mass or vehicle motion - but this will not inform you about road condition
    • Typical data capture using linear potentiometers and even non mechanical laser, optical, ultrasonic distance capture methods are fragile
  • AWS Snow devices provide edge processing capability especially for larger remote facilities like forward operating bases, field hospitals, mine sites, etc. They also integrate well with satcom ground terminals to provide cost effective remote data aggregation and processing https://aws.amazon.com/blogs/publicsector/enhance-operational-agility-decision-advantage-aws-snowball-edge/