Simulation of autonomous intersection management
Michał Tomański
Promotor: dr hab. inż. Jarosław Wąs
Overview
Why?
What?
How?
Current situaton
So, what's the problem?
The infrastructure is not ready
Classic intersections won't be effective
Parking places could be more 'intelligent'
They need a huge improvement!
Goal
To create an intersection management system for autonomous vehicles
Investigate the system under simulation
Also simulate traditional approaches (traffic lights, roundabout, intersection with priorities)
Compare the results of simulation and choose a winner
Autonomous Intersection Management
A vehicle approaching the intersection sends a notice to the managing server
The notice contains incoming direction and desired direction of the vehicle
The notice is added to the queue managed by the server
The manager keeps track of the intersection's area usage in a time slot
Based on the usage, manager sends back an approval of crossing or a rejection
Approval
The manager is responsible for calculating the route of the vehicle
The route cannot interfere with already allocated routes
Constant speed and constant angular velocity is assumed
The car is required to slow down to the given speed
Additional things to consider
Common notice protocol
Safety measures - safe distance, speed
Protocol of retrying in case of network failure
Behavior of vehicles in case of software failure
Brief specification of hardware needed for both server and vehicles
Improvements
Nonlinear speed, nonlinear angular velocity
Parametrised vehicles
Priorities - emergency vehicles, buses, bus lanes
Priority algorithms - first come first served, first route availabe first served
Time optimisation - one vehicle optimisation, entire crossing throughput optimisation - Altruism!
Fuel consumption optimisation
Audit logs - legal concerns
Technical stuff
Language: scala
Web framework: Play
Concurrency mechanism: Akka actors
Messaging system and persistence: Kafka
Front-end: ReactJS with typescript (just for simulation)
Thanks
Made with Slides.com