Cornell University · IDS Lab

Scaled City
Experiments

I conduct experiments in the Information and Decision Science Lab Scaled Smart City (IDS3C), a 1:25 scale robotic testbed located in Hollister Hall at Cornell University. The testbed replicates real-world urban traffic scenarios in a controlled environment, enabling us to validate control algorithms beyond simulation before transitioning to full-scale deployment.

My experimental work spans coordinating autonomous ground vehicles with quadcopters (Crazyflies), testing CAV interactions with human-driven vehicles (HDVs) through driving emulators, and validating signal-free intersection coordination — all using real-time optimal control frameworks with VICON motion capture localization.

1:25 Scale Testbed VICON Motion Capture Crazyflie Quadcopters CAV · HDV Optimal Control Hollister Hall · Cornell

IDS3C Testbed
Experiment 01

Signal-Free Intersection Coordination

Autonomous vehicles navigate a multi-lane intersection without traffic signals, coordinating through a decentralized optimal control framework. Each vehicle computes its trajectory in real time to cross the intersection safely and efficiently.

The experiment demonstrates energy-optimal merging and crossing maneuvers with provable safety guarantees, bridging the gap between simulation results and physical deployment.

Signal-Free Intersection Optimal Control Multi-Vehicle
IDS3C Testbed
Experiment 02

CAV Interaction with Human-Driven Vehicles

Connected and automated vehicles (CAVs) coordinate alongside human-driven vehicles (HDVs) simulated through driving emulators. The framework handles the uncertainty of human behavior while maintaining safety and throughput.

This mixed-traffic scenario is critical for real-world deployment, where full automation penetration is unlikely in the near term.

Mixed Traffic HDV Emulator CAV Human Factors
IDS3C Testbed
Experiment 03

Ground–Aerial Vehicle Coordination

Coordinating autonomous ground vehicles with Crazyflie quadcopters in a joint search-and-rescue mission scenario. Ground vehicles and aerial drones operate as a heterogeneous team, each assigned roles based on their sensing and mobility capabilities.

The multi-agent decision-making framework allocates tasks dynamically and re-plans in real time as the environment evolves, validated end-to-end in the scaled testbed.

Multi-Agent Crazyflie UAV Heterogeneous Teams Search & Rescue
IDS3C Testbed
Experiment 04

IDM Car Following with Yield Sign

Vehicles following the Intelligent Driver Model (IDM) respond to yield sign scenarios, testing the interaction between rule-based behavior and optimal control policies in a structured urban environment.

This experiment studies how IDM-governed vehicles integrate with infrastructure cues, informing the design of controllers that must coexist with legacy driving behavior.

IDM Car Following Yield Sign Urban Traffic