Expressive Real-Time Intersection Scheduling
Authors: Rick Goldstein, Stephen Smith
AAAI 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We ran evaluations of the three methods on the Simulation of Urban MObility (SUMO) (Krajzewicz et al. 2012). SUMO is a microscopic traffic simulator that simulates the dynamics of vehicles and traffic light controllers. We interface with SUMO through Python, using the Traffic Control Interface (Tra CI) (Wegener et al. 2008). Tra CI provides real-time data about vehicle locations and allows external control of traffic decisions. All simulations run for an hour of simulated time. For each simulation, we report average time loss, per vehicle. To eliminate the effects of simulation start up and termination, we only report the time loss of vehicles arriving within the middle 40 minutes. Results for a given experiment are averaged across ten simulation runs with different random seeds. Identical random seeds are used when comparing the three methods to generate identical traffic arrivals. Vehicle arrivals are modelled as a Poisson process where the average arrival rate is set according to the desired level of congestion. We present a comparison of these methods on two separate networks, a discussion of statistical significance, and a timing analysis on components of the Scheduling Module. |
| Researcher Affiliation | Academia | Rick Goldstein The Robotics Institute Carnegie Mellon University rgoldste@cs.cmu.edu Stephen F. Smith The Robotics Institute Carnegie Mellon University sfs@cs.cmu.edu |
| Pseudocode | Yes | Algorithm 1 Calculate State Transitions |
| Open Source Code | No | The paper does not provide any statement or link indicating that the source code for ERIS or the described methodology is publicly available. |
| Open Datasets | No | The paper uses the SUMO simulator and states, "Vehicle arrivals are modelled as a Poisson process where the average arrival rate is set according to the desired level of congestion." This indicates data was generated for simulation rather than using a publicly available dataset, and no access information is provided for any generated data. |
| Dataset Splits | No | The paper describes how simulations were run and results were reported (e.g., "All simulations run for an hour of simulated time", "only report the time loss of vehicles arriving within the middle 40 minutes"), but it does not specify any explicit training, validation, or test dataset splits or cross-validation methodology. |
| Hardware Specification | No | The paper mentions running simulations on SUMO but does not provide any specific details about the hardware (e.g., CPU, GPU, memory, cloud instance type) used for these experiments. |
| Software Dependencies | No | The paper states that experiments were run on "Simulation of Urban MObility (SUMO) (Krajzewicz et al. 2012)" and that they "interface with SUMO through Python, using the Traffic Control Interface (Tra CI) (Wegener et al. 2008)". While the software names are provided, specific version numbers for Python, SUMO, or Tra CI are not mentioned, which is required for reproducibility. |
| Experiment Setup | Yes | We assume that all intersections are four-sided intersections controlled by a dual ring barrier controller. ... Green movements must satisfy prespecified minimum and maximum timing limits. Similarly, we require yellow and red clearances with fixed lengths to occur between successive green movements for safety. ... The dual ring barrier controller allows left turn movements to end at separate times; this increases scheduling flexibility but requires keeping track of green movements and their respective start times separately. ... In our dual ring controller, left turn movements precede straight movements and left turn movements can be skipped if no vehicles are present. Straight movements cannot be skipped. Currently, we allow only protected movements at our intersections. ... fixed cut-off of 3 seconds... minimum green time requirement of 10 seconds. Yellow and red clearances are a combined 5 seconds. ... All simulations run for an hour of simulated time. For each simulation, we report average time loss, per vehicle. ... Results for a given experiment are averaged across ten simulation runs with different random seeds. Identical random seeds are used when comparing the three methods to generate identical traffic arrivals. Vehicle arrivals are modelled as a Poisson process where the average arrival rate is set according to the desired level of congestion. |