MMDP: The Key To Smarter Bike And Scooter Sharing
MMDP
Quantum Annealing Improves Urban Transit Micro-Mobility. As municipalities globally aspire toward greener transportation, bike and scooter sharing have become practical challenges. The “stochastic” and “highly dynamic” nature of metropolitan demand, where client habits change quickly and cars need to be reassigned regularly, has historically plagued these systems. However, a groundbreaking study by Tohoku University, Honda R&D, and Sigma-i Co., Ltd. indicates that quantum physics can solve these complicated urban difficulties.
Beyond Classical Logistics
The Vehicle Routing Problem (VRP), a “NP-hard” combinatorial optimization, has dominated logistics for decades. VRP, including Capacitated VRP (CVRP) and VRP with Time Windows (VRPTW), has been successful in global shipping and delivery, but it is becoming unsuitable for micro-mobility.
Micro-mobility systems use autonomous or semi-autonomous single-passenger cars that must move to meet demand, unlike delivery trucks with fixed routes. Due to demand volatility, long-term route planning is less relevant. Academics Takeru Goto and Masayuki Ohzeki proposed the Micro-Mobility Dispatch Problem to tackle this.
Synergy between quantum and bayesian
Bayesian analysis of previous usage data underpins this novel technique. The system may optimally distribute idle vehicles over several charging and standby stations by evaluating client arrival patterns and destination decisions.
The researchers used a Quadratic Unconstrained Binary Optimization (QUBO) model for quantum solver compatibility. Unlike typical heuristics that focus on the nearest vehicle to a consumer, QUBO considers the entire network.
This concept uses complex mathematical “Hamiltonians” to limit and reduce costs. For instance:
HA0 assigns each vehicle to one customer or station. Every customer request is assigned to one vehicle via HA1. To reduce vehicle travel time between destinations and targets, HB0 includes the overall journey time cost. Based on prior data, HB1 encourages automobiles to concentrate in high-consumer-appearance areas.
Powerful D-Wave Advantage
The team used the commercial quantum annealer D-Wave Advantage to test their theory. Quantum Annealing (QA) uses quantum physics to do difficult optimization computations ten times faster than computers.
Reverse Annealing was a study highlight. RA refines solutions by starting from a “high-quality initial state” and carefully altering the transverse field to better explore the solution space than forward annealing. This strategy greatly improved solution quality, allowing the quantum solver to outperform the Gurobi Optimizer, a high-end classical solver, under certain conditions.
Dynamic vs. Static Methods
The study examined two historical data addition methods:
The Dynamic Approach: Real-time car placements reduce consumer wait times. While it provides the best service, it often increases fleet travel time.
The Static Approach: This technique drives cars to high-frequency areas using statistical data without real-time vehicle tracking. A “balanced improvement” in service quality without dramatically increasing journey time.
No matter the request frequency, the dynamic method outperformed greedy algorithms in key service measures in experiments.
The Value of Calibration
Quantum formulation success depends on variable balance. Adjusting the weight ratio between imminent travel costs (B0) and historical demand promise (B1) is crucial, the researchers discovered. They found that B1 =0.3 and B0 =0.1 were ideal through empirical testing. The customer-assignment phrase (HA1) was crucial, and eliminating it significantly reduced performance, according to a “ablation study”.
Toward Quantum Future
The findings are promising, but the authors caution. The existing model is based on probability distribution approximations and has a “cyclical interplay” in which operational settings affect performance metrics and subsequent data.
Future studies will likely focus on this feedback loop's stability and the use of model-free estimate methods like neural networks to improve dispatch logic. There is also interest in a “hybrid dynamic-static scheme” to balance energy use and service quality.
The effects on urban transit are considerable. The ability to “boost fleet utilization and reduce wait times” gives micro-mobility providers an advantage in cities that want to reduce congestion and carbon emissions. By using quantum annealing, future urban transit may be driverless, electric, and quantum-optimized.