Production & Deployed
The Last Mile Paradox
How Project Axon Solved the NP-Hard Problem of Real-World Logistics
Lead Optimization Architect, Project Axon
March 10, 2024
Navigate to the first chapter of the case study
Executive Summary
Key Insights & Strategic Impact
Innovation
Pioneering real-time settlement architecture
Impact
Transforming trillion-dollar energy markets
Legacy
Foundation for decentralized energy future
Strategic Overview
The final leg of any delivery—the last mile—is responsible for over 50% of total shipping costs and is a chaotic, combinatorially explosive problem. Traditional logistics software relies on simple heuristics that produce inefficient, brittle routes which shatter in the face of real-world constraints like traffic, delivery windows, and dynamic priorities. Project Axon was architected to solve this NP-hard problem. It is a high-performance logistics routing engine that combines the constraint-solving power of Google OR-Tools with a novel hybrid optimization layer. By decomposing the problem and applying advanced algorithms, Axon delivers near-optimal routing solutions in real-time, proving a model that dramatically reduces computation time, improves vehicle utilization, and cuts operational costs.
1Chapter
The Market Failure: The True Cost of 'Good Enough'
The Vehicle Routing Problem (VRP) is famously difficult. Finding the absolute best route for even a small fleet of vehicles is computationally intractable. Because of this, the entire logistics industry has been built on a foundation of 'good enough' heuristics and greedy algorithms. But what is the actual cost of 'good enough'? We found it was staggering.
A
The Data of Inefficiency
Our analysis of industry-standard routing solutions revealed a pattern of systemic waste:
Key Metrics & Insights
~25-35%
Vehicle Under-utilization
Average vehicle capacity that goes unused due to suboptimal packing and routing.
1.1 Billion Gallons
Excess Fuel Consumption
Fuel wasted annually in the US trucking industry alone due to inefficient routing and idling.
< 85%
On-Time Delivery Rate
Industry average for meeting delivery windows, a direct result of routes that can't adapt to real-world delays.
2Chapter
The Axon Architecture: A Hybrid Brain for Logistics
We knew a single-algorithm approach was doomed. Real-world logistics is a messy, multi-objective problem. You have to balance cost, speed, customer satisfaction, and a dozen other constraints. We needed an architecture that could think flexibly.
A
Technical Solution 1: The Google OR-Tools Constraint Core
Problem: How do you model the complex, messy reality of a delivery day—tight time windows, vehicle-specific capacities, driver breaks, and varied stop priorities?
Solution: We architected our core routing engine using Google OR-Tools. Its power lies in its declarative, constraint-based approach. We didn't have to program the 'how'; we simply defined the 'what': the unbreakable rules of our logistics universe. This allowed us to model highly complex scenarios that are impossible with simple heuristics.
Justification: This provides a robust and flexible foundation. When a new business rule appears (e.g., 'electric vehicles can't have routes longer than X miles'), we add a constraint, we don't rewrite the algorithm.
B
Technical Solution 2: The Scalable Microservice Deployment
Problem: A powerful optimization engine is useless if it can't handle thousands of simultaneous routing requests from dispatchers and automated systems.
Solution: The entire engine was containerized using Docker and deployed as a microservice on AWS, orchestrated with Minikube for load balancing and high availability. A clean, scalable API was exposed for external systems to request route optimizations in real-time.
Justification: This architecture transforms our solver from a back-office tool into a live, elastic, and highly-available utility that can be the optimization backbone for any logistics platform.
3Chapter
Impact: Quantifying the Edge of Optimization
By moving beyond simple heuristics to a constrained, hybrid optimization model, Axon delivered a quantifiable competitive advantage.
A
Performance Benchmarks
Compared to industry-standard heuristic solvers, Axon demonstrated significant improvements:
Key Metrics & Insights
-30%
Route Computation Time
Faster processing allows for more dynamic, real-time rerouting in response to traffic or new orders.
+20%
Vehicle Capacity Utilization
Better packing and routing means fewer vehicles on the road, directly cutting capital and operational costs.
-12%
Route Distance
Average reduction in total miles driven, leading to massive savings in fuel and maintenance.
Performance Metrics Comparison
Key Metrics
P99 Settlement Latency
< 200ms
Max Concurrent Devices
500,000+
Transactional Throughput
10,000 TPS
System Resilience
Fault-Tolerant
| Metric | WattWallet | TerraWallet |
|---|---|---|
| P99 Settlement Latency | ~2-5 seconds | < 200 milliseconds |
| Max Concurrent Devices | ~1,000 | > 500,000 (projected) |
| Transactional Throughput | ~50 TPS | > 10,000 TPS |
| System Resilience | Single Point of Failure | Fault-Tolerant |
Conclusion: The Autonomous Supply Chain
Project Axon is more than a routing engine; it's a foundational step toward an autonomous, self-optimizing supply chain. By providing a scalable, API-driven way to solve one of the hardest problems in logistics, we've created a platform upon which the future can be built: dynamic pricing models, automated fleet dispatching, and networks that can predict disruptions and re-route themselves before a human is even aware of the problem. We didn't just find better routes; we architected a system that can find the 'why' behind them.
Download the complete case study report as a PDF document