Autonomous on-Demand Shuttles for First Mile-Last Mile Connectivity: Design, Optimization, and Impact Assessment

Sudipta Roy, Gabriel Dadashev, Lampros Yfantis, Bat-hen Nahmias–Biran, Samiul Hasan

TL;DR

Researchers have developed a simulation-based controller that optimizes autonomous on-demand shuttles (AODS) for "First-Mile Last-Mile" connectivity in suburban areas. The study demonstrates that a small fleet of just three shuttles can serve 80% of demand with wait times under four minutes, all while having a negligible impact on existing traffic flow.

What Problem Does This Paper Address?

Public transit often fails in the "sprawling" suburbs. While trains or express buses might get you to a central hub, the "First-Mile Last-Mile" (FMLM) problem—the gap between your front door and that hub—remains a major barrier. Traditional fixed-route buses in these low-density areas are expensive and inefficient, leading most people to rely on personal vehicles.

Autonomous on-Demand Shuttles (AODS) offer a promising middle ground. However, deploying them effectively requires solving a complex optimization puzzle: How do you route them in real-time based on fluctuating demand? How do you manage charging without depleting the fleet? And crucially, will these slow-moving shuttles (typically capped at 15–25 mph) create a nightmare for regular drivers? This paper tackles these questions by integrating realistic traffic modeling with fleet optimization.

Methods

The researchers built a custom AODS controller using the Aimsun Ride API and Python. The framework stands out by combining three distinct layers:

  1. Stop Optimization: They used K-means clustering to determine the most efficient locations for shuttle stops, ensuring that 96% of residents are within a 6-minute walk of a stop.
  2. Routing & Assignment: The controller uses a greedy insertion heuristic. When a request comes in, the system checks all active shuttles and finds the "least costly" way to insert the new pickup/drop-off into an existing schedule without violating maximum wait time thresholds.
  3. Dynamic Traffic Assignment (DTA): Unlike simpler models that assume constant speeds, this study uses mesoscopic simulation. This accounts for how real-world traffic congestion affects shuttle travel times and, conversely, how the shuttles affect the cars around them.
  4. Charging Logic: The model simulates DC Level 3 (50 kW) chargers, implementing a "critical limit" logic where shuttles automatically go offline to recharge when battery levels drop below 15 kWh.

Datasets

The study is grounded in a high-fidelity digital twin of Lake Nona, Orlando, Florida, a fast-growing suburban community.

  • Demand Data: Initial mobility patterns were sourced from Teralytics Studio (anonymized GPS, census, and employment data).
  • Traffic Calibration: The simulation was calibrated and validated using real-world road detector data from the Regional Integrated Transportation Information System (RITIS), achieving a high GEH score (a standard metric for traffic model accuracy) of 3.94.
  • Transit Ratios: The researchers used the National Household Travel Survey (NHTS) to estimate that roughly 1% of total trips currently act as transit riders, with a "futuristic" scenario set at 2%.

Key Insights and Ideas

  • The Power of Three: In the Lake Nona case study, a fleet of just 3 shuttles reached a "sweet spot," serving 80% of requests. Adding more shuttles provided diminishing returns for rider satisfaction while increasing operational costs.
  • The "Slow Move" Myth: One of the most significant findings is that low-speed shuttles (15 mph) have a negligible impact on regular traffic (25 mph roads). The mean speed of human-driven vehicles decreased by only 0.1 mph. This suggests that suburban residential roads are ideal "laboratories" for early AV deployment.
  • Charging is the Bottleneck: The study highlights that "inactive waiting time" for a charger is a silent killer of fleet efficiency. If charging infrastructure isn't optimized, you need a significantly larger fleet just to maintain the same level of service.
  • Efficiency over Walking: Even with the smallest fleet tested, the total trip time (wait + ride) was consistently 2.5 times faster than the walking alternative.

Key Results

  • Wait Times: Average waiting times remained under 4 minutes for the 3-shuttle fleet.
  • Service Rate: Approximately 80% of FMLM requests were successfully served on-demand.
  • Coverage: 15 strategically placed stops covered 96% of the parcels in the study area.
  • Traffic Impact: The delay-to-travel-time ratio for regular vehicles remained largely unchanged (around 1.2%), proving the shuttles don't cause significant congestion.

Limitations and Future Work

While the results are promising, the authors acknowledge several areas for further exploration:

  • Geographic Specificity: The findings are specific to suburban Lake Nona. Urban environments with higher density and different speed limits might yield different traffic impact results.
  • Static vs. Dynamic Stops: The current model uses fixed stops. Future work could investigate "door-to-door" service or dynamic stop relocation.
  • Advanced Charging: The study suggests investigating quasi-dynamic wireless charging (charging while the vehicle is briefly stopped or moving) to further reduce the "inactive" time that currently hampers fleet performance.
  • Relocation Strategies: The controller does not currently "rebalance" the fleet (moving empty shuttles to high-demand areas before requests arrive), which could further optimize wait times.

Summary updated

Abstract

The First-Mile Last-Mile (FMLM) connectivity is crucial for improving public transit accessibility and efficiency, particularly in sprawling suburban regions where traditional fixed-route transit systems are often inadequate. Autonomous on-Demand Shuttles (AODS) hold a promising option for FMLM connections due to their cost-effectiveness and improved safety features, thereby enhancing user convenience and reducing reliance on personal vehicles. A critical issue in AODS service design is the optimization of travel paths, for which realistic traffic network assignment combined with optimal routing offers a viable solution. In this study, we have designed an AODS controller that integrates a mesoscopic simulation-based dynamic traffic assignment model with a greedy insertion heuristics approach to optimize the travel routes of the shuttles. The controller also considers the charging infrastructure/strategies and the impact of the shuttles on regular traffic flow for routes and fleet-size planning. The controller is implemented in Aimsun traffic simulator considering Lake Nona in Orlando, Florida as a case study. We show that, under the present demand based on 1% of total trips as transit riders, a fleet of 3 autonomous shuttles can serve about 80% of FMLM trip requests on-demand basis with an average waiting time below 4 minutes. Additional power sources have significant effect on service quality as the inactive waiting time for charging would increase the fleet size. We also show that low-speed autonomous shuttles would have negligible impact on regular vehicle flow, making them suitable for suburban areas. These findings have important implications for sustainable urban planning and public transit operations.
BibTeX
@misc{roy2024autonomous,
  author = {Sudipta Roy and Gabriel Dadashev and Lampros Yfantis and Bat-hen Nahmias–Biran and Samiul Hasan},
  title = {{Autonomous on-Demand Shuttles for First Mile-Last Mile Connectivity: Design, Optimization, and Impact Assessment}},
  year = {2024},
  eprint = {2408.07872},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2408.07872},
}
Figure