3 Auction-Based Traffic Projects
Project 1 of 3: Auction-Based Traffic Signal Control
Keywords: resource allocation, auction-based systems, V2I & V2V communications
Goal: A control algorithm capable of accommodating user inputs and needs.
Problems: Each driver wants to reach his/her destination in the shortest time. Yet, no trips are the same: ambulance, delivery truck, transit vehicles, and a trip to Walmart. Existing algorithms can’t accommodate individual desires.
Solution: Auction-based control algorithms that bridge the gap
Results:
- Improved overall performance by 20-25% compared to the state-of-the-art algorithm
- Ability to provide better service for “important” trips without causing extra delay for “normal” trips (dark and light blue graphs)
Commercial Application: “Route Caste” framework in Surtrac is built on these ideas.
Technical Contribution
- This is the first successful framework that employed auction-based ideas to signal control problem without violating any real-world constraints.
- This research is carried out as a part of my doctoral work
Project 2 of 3: Auction-Based Fleet Scheduling
Keywords: resource allocation, auction-based systems, paratransit, dial-a-ride problem
Goal: A real-time task planning algorithm for fleet scheduling
Problems:
- Computational complexity is NP-Hard
- Large-scale problem (40 vehicles, 1500+ trips)
- Most existing optimization models can only solve smaller problems (# of vehicles less than 10), except the Generalized Task Swap (GTS) algorithm
Solution: Auction-based task planner framework that operates in real-time
Results:
- Improved overall performance by 20-25% compared to the state-of-the-art algorithm
- Ability to provide better service for “important” trips without causing extra delay for “normal” trips (dark and light blue graphs)
Commercial Application: The efficacy is tested on Access paratransit data
Technical Contribution
- Oversaw algorithmic development
- Oversaw scheduling simulator development
Project 3 of 3: Auction-Based Multi-Robot Scheduling
Keywords: resource allocation, auction-based systems, drone scheduling, ISR missions
Goal: A real-time task planning algorithm for fleet scheduling
Problem:
An autonomous decision-support system framework for ISR missions
Requirements:
- Autonomously process incoming requests and rank-order them based on operational constraints
- Identify and optimally allocate appropriate resources to successfully fulfill the maximum number of requests without violating constraints
- Ability to operate in real-time
Solution:
- Two layered scheduler – for scheduling reconnaissance and assault drones
- Auction-based task planner
Results:
The auction-based task scheduler was able to improve resource utilization between [6, 40] percent by 40% (best case), 6%(worst case), 25% (on average)
Commercial Application: The efficacy is tested on Access paratransit data
