Planning Under Uncertainty for a Last-Mile Delivery Fleet
Overview
What this challenge is about.
Build a simulator of the 50-block area with stochastic travel times conditioned on weather and time-of-day. Implement value iteration (for a small state space), MCTS (Monte Carlo Tree Search), and the current heuristic. Evaluate on 200 simulated dispatches across 7 days x 4 weather scenarios. Measure expected delivery time + 95th-percentile time. Deliver a 4-page memo recommending one method (or a hybrid) for the next pilot.
The Brief
What you'll do, and what you'll demonstrate.
Pick the best planning approach for stochastic last-mile dispatch on expected and tail delivery time under realistic weather variability.
Earning criteria — what you'll demonstrate
- Model a dispatch problem as a Markov Decision Process
- Implement and tune Monte Carlo Tree Search on a realistic state space
- Compare planners on expected and tail metrics (not just average)
- Translate stochastic planning results into business recommendations
Program Fit
Where this fits in your program.
Sharpens the same skills your degree expects you to demonstrate.
Skills
Skills you'll demonstrate.
Each one shows up on your verified credential.
Careers
Roles this prepares you for.
Real titles. Real skill bridges. Pick the one closest to your trajectory.
ML Researcher
MDP modeling + MCTS benchmarking on a real ops problem is the experimental work ML researchers in industrial AI do regularly.
This challenge sharpens
- markov-decision-processes
- monte-carlo-tree-search
- benchmarking
AI Engineer
Stochastic planning prototypes paired with simulator infrastructure is high-leverage AI engineering work at logistics-AI startups.
This challenge sharpens
- simulation
- planning-under-uncertainty
- monte-carlo-tree-search
Applied AI Scientist
Tail-aware reporting and pilot-recommendation framing is the daily output of an applied AI scientist at a delivery company.
This challenge sharpens
- planning-under-uncertainty
- benchmarking
- simulation