Mobile Grocery Retail Network Optimisation
Waterloo–Kitchener, Ontario, Canada · Independent Synthetic Case Study
This independent case study illustrates integrated network planning and multi-period scheduling for mobile grocery retail operations. It explores facility location and allocation, resource planning, multi-objective mathematical optimisation, solution algorithms, computational benchmarking, and interactive decision-support visualisations.
Work Demonstrated
1. Data Engineering
A Python-based data preparation workflow validates and transforms synthetic customer-zone, candidate mobile grocery location, and temporal demand inputs into optimisation-ready datasets.
Overview
Operational datasets may contain incomplete, inconsistent, or invalid records that must be addressed before reliable planning decisions can be made. This stage validates postal-code and coordinate information, identifies invalid entries, and extracts temporal demand patterns to produce structured inputs for optimisation.
Location records containing customer zones and candidate mobile grocery locations.
Historical demand activity summarised across multiple temporal resolutions.
Profiles represent aggregated historical demand activity within geographic customer zones, rather than demand frequency for individual customers.
2. Road-Network Distance Modelling
Developed an automated routing pipeline using OSRM APIs and Docker to construct origin-destination road-network distance matrices for customer access network optimisation.
Overview
Straight-line measures such as Euclidean and Haversine distance are straightforward to calculate, but they do not represent actual road travel. This stage uses road-network routing to obtain travel distances and estimated driving times between customer zones and candidate mobile grocery locations.
Road-network comparison
3. Mathematical Optimisation
Formulated an integrated multi-objective optimisation framework encompassing mobile grocery location selection, customer assignment, resource allocation, and weekly operational scheduling.
Overview
The problem is formulated as an integrated multi-objective optimisation framework structured across two interconnected planning levels. The strategic planning level is formulated as a single-source capacitated facility location problem (SSCFLP), while the operational planning level is formulated as a multi-period mobile grocery scheduling problem, using the resulting strategic network configuration to determine detailed weekly service schedules.
- Mobile grocery location selection
- Customer-zone assignment
Customer zones and candidate mobile grocery locations.
Facility Location Problem
Determines facility opening decisions and assigns each customer zone to one selected facility.
Open grocery locations and single-source customer assignments.
The figure communicates the decision structure of the strategic optimisation layer without disclosing operational data, model parameters, constraints, or mathematical formulations.
- Facility operating days
- Service-session timing
Selected grocery locations are translated into daily service schedules across a seven-day planning horizon.
The timetable communicates the structure of the operational scheduling layer without exposing schedules, staffing information, capacities, or model constraints.
• Unmet demand
• Travel distance
• Operational cost
as competing objectives
Pareto front approximation using the augmented ε-constraint method, solved with Gurobi (12.0.0)
Methodological details of the proposed algorithms are intentionally omitted to protect ongoing research and intellectual property.
4. Computational Experiments
Computational experiments evaluate the optimisation methods across operational performance, computational efficiency, and multi-objective solution quality.
Overview
The proposed solution algorithms are compared with a commercial exact optimisation solver using illustrative benchmark results. The evaluation considers unmet demand, computational runtime, and multi-objective solution quality through inverted generational distance (IGD+) and hypervolume (HV).
Primary operational performance measure, capturing unmet customer demand.
Box plots summarise the distribution of unmet demand across repeated runs for the selected benchmark instance.
Boxes show the interquartile range and median; whiskers extend to the most extreme non-outlier observations. The × marker shows the mean.
Evaluates the computational effort required by each optimisation method.
Mean runtime across repeated runs is compared across Medium and Large benchmark datasets.
Runtime is shown on a logarithmic scale to accommodate the substantial differences in computational time across solution methods.
Runtime is shown on a logarithmic scale to accommodate the substantial differences in computational time across solution methods.
Evaluates multi-objective solution quality using complementary convergence and coverage indicators.
IGD+ and hypervolume (HV) summarise the quality of the obtained non-dominated solution sets across repeated runs.
Both indicators use a common reference approximation set constructed from the union of non-dominated heuristic and Gurobi solutions.
Boxes summarise repeated heuristic runs; the dashed horizontal line shows the corresponding Gurobi benchmark indicator value.
5. Decision-Support Platform
Developed and deployed a browser-based decision-support platform integrating data preparation, configurable optimisation, network visualisation, operational scheduling, scenario analysis, and performance analytics.
Overview
This interactive decision-support platform brings together data preparation, network planning, scenario analysis, weekly scheduling, and performance evaluation in a single workflow. The independent synthetic demonstration below illustrates how these planning decisions can be explored through maps, schedules, and performance visualisations.
Planning Capabilities
The platform integrates data preparation, mathematical optimisation, network visualisation, scenario analysis, and performance evaluation within a unified planning environment.
Mobile Grocery Planning Platform Demonstration
Click the button to follow the planning workflow step by step inside one interactive application.