Fleet Management and Optimisation - Industrial Placement Presentation

Lorenzo Paoliani Industrial Placement 2016MEng Computing | Imperial College London

Fleet Management and Optimisationcode that delivers

Pie is a solution for logistics and transportation companies to manage their

vehicles and power their operations.

PLANNING TRACKING ROUTING

Manage drivers and fleet

Register customer orders

Track vehicles on the roadApp for drivers

Route vehicles from A to B

Route around restrictions for large vehicles, road

closures, etc.

Apple Pie

Over the course of my placement I focused on 3 areas

UI UX DXUser

InterfaceUser

ExperienceDeveloper Experience

Storybook

DX

Storybook• Separates “pure” view components from

the main app • Allows to describe the intent behind a

component by describing a story of its possible rendering states

• Distraction free environment • Quickly iterate • Communicate with the design team • Track use cases, error and loading states

FiltersUI UX

Filters• Logic and UI to filter deeply nested data • Filters pile on top of each other • Recursively descends into an entity

checking whether the current piece of information is hidden or visible

• At every level, after visiting the children nodes, the parent decides its status

• This allows to mark every node as one of VISIBLE / DISABLED / HIDDEN

Cuttlefisha network wide route optimisation engine

UX DX

The Problem

The Problem

• 100+ locations to dispatch vehicles • Thousands of vehicles • Pick up freight from 300+ locations all over the UK every day • Sort the deliveries and send them towards the right regional

depot • Must arrange a plan to fulfil all the orders

The Problem• 130+ locations to dispatch vehicles • Thousands of vehicles • Pick up freight from 300+ locations all over the UK every day • Sort the deliveries and send them towards the right regional

depot • Must arrange a plan to fulfil all the orders

Right now, this is done in an office, by hand, every day.

Engine

based on a 2011 paper:An Iterated Local Search heuristic for the Heterogeneous Fleet Vehicle Routing Problem• Defines the HFVRP and its subcategories • 2 hard problems in computer science

• Travelling Salesman Problem • Bin Packing Problem

• Searches solutions iteratively through a small subset of similar solutions from the solution space

• Uses random perturbation of candidate solutions to escape local minima

• Any solution - even no optimisation! - is better than the current state of the long haul logistics

Penna, P.H.V., Subramanian, A. & Ochi, L.S. J Heuristics (2013) 19: 201. doi:10.1007/s10732-011-9186-y

API + Solver• Exposes an API to build and solve a

Heterogeneous Fleet Vehicle Routing Problem

• Input: a set of dispatching locations and a set of pickup jobs

• Output: a fulfilment plan that connects jobs and dispatchers

• Handles pickup time, service time, volume, and weight constraints

• Selects best vehicle type to service a route

App

Build a Problem

Plot the Solution

Technology

graphhopper/jspritan open source implementation of the algorithms described in

the HFVRP paper

Statically typed programming language

for the JVM

Engine App

Lorenzo Paoliani [email protected]

Industrial Placement 2016MEng Computing | Imperial College London

Thanks!