IoT | Machine Learning | AI | Blockchain
Usecase Prelude
The future is embracing Autonomous Machine Economy where Machines Co-exist and interacts with Humans withelan, using technology advancements in the form of IoT, AI, Machine Learning and Blockchain
Introduction to Our Solution
Imagine a fleet of autonomous cars, able to cluster themselves, align themselves based on their performance and map themselves automatically on to areas of High and Low Demands for Shared Fleet Service, without HUMAN
INTERVENTION
Problem Statement
In a world of Shared Fleets, is It possible to imagine that an Autonomous Car can transport humans from passenger pick up areas of High Demand or Low Demand by self evaluating its own performance ? Based on the vehicle telematics details, car trip data, maintenance pit stops, charing stops etc. can an Autonomous and a Decentralised AI engine enable an autonomous car to self evaluate its own performance and auto allocate itself in areas of high or low demands ?
Solution Overview
An amalgamation of unique tech stacks can make this a reality – AI-Priori along with Spherity, worked on an ensemble experiment correlation IoT, ML, AI with a Blockchain based DApp.
Solution Stack – Digital identities of the Fleets, Passengers and other metrics were verified through Spherity decentralised digital twin protocol. The identities created were stored in the backend using IPFS and BigchainDB services. The architecture was modelled on Ethereum to invoke Smart Contracts controlling individual fleet transactions. Each transaction processed was stored back into the BDB Blockchain source. The frontend invoking of transaction requests are initiated through a Blockstack powered DApp.
Dual Machine Learning – Data Modeling
Car Performance – A Data of 100 Autonomous Tesla Cars were taken as a preliminary data set from Uber’s Data Movement Library. Raw data was chiselled through automatic web scrapping to come up with key autonomous car performance criteria such as: Car Mileage, Trip Data Points, Maintenance Pitstops, Charging Pitstops, etc. Based on the Chiselled Data, customised ML modeling was done to determine attributes of Highly Performing cars and lowly performing cars.
Computing Areas of High and Low Demand – A geographic 10 KM radius was mapped for the city of Paris, France, for the 100 cars trip detail report. Based on the pick up and drop points and number of trips made etc., we devised a customised ML model based predictive logic to calculate areas of High and Low Demands, on specific times in a day/month/year etc.
Decentralised AI Automation – Combining both the data models above, a logic was defined to invoke a smart contract notification for cars with high performance to auto allocate itself to the area of high demand and lowly performing cars were to auto allocate to areas of Low demand. From an architecture.
DApp Front End – To render end to end exposure, we have built an agile and scalable DApp powered by Blockchain tied to Ethereum and BDB/IPFS backends.The workflow gets triggered ideally with a source file upload, carrying Fleet level aggregated data points. The DApp enables the fleets to trigger autonomous transactions at an individual fleet level .The DApp then invokes the ML/AI engine to rationalise the Demand and Fleet performance metrics and automatically allocates itself to high demand areas in anticipation of Passenger Fleet request with minimal wait times for the shared fleet user.
