Predicting Vehicular Travel Times by Modeling Heterogeneous Influences Between Arterial Roads
Authors: Avinash Achar, Venkatesh Sarangan, Rohith Regikumar, Anand Sivasubramaniam
AAAI 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Using synthetic and real world data traces we demonstrate the superior performance of the proposed method under different traffic conditions. We test the usefulness of our approach on both synthetic data and real-world probe vehicle data obtained from the cities of (i) Porto, Portugal and (ii) San Fransisco. On synthetic data, relative absolute prediction error reduces by as much as 70% under the proposed method in the worst case. On real world data traces from Porto and San Fransisco, the proposed approach performs up to 14.6% and 16.8% better respectively than existing approaches in the worst case. |
| Researcher Affiliation | Collaboration | Avinash Achar, Venkatesh Sarangan Rohith Regikumar TCS Research, IIT Madras Research Park Chennai 600113, INDIA. Anand Sivasubramaniam Dept. of Comp. Sci & Eng., Pennsylvania State University, State College, PA 16802, USA. |
| Pseudocode | Yes | Algorithm 1: Compute expected travel time of an aribitrary length query route |
| Open Source Code | No | The paper references its own technical report on arXiv (Achar et al. 2017. Predicting vehicular travel times by modeling heterogeneous influences between arterial roads. Technical Report ar Xiv: 1711.05767 [cs.AI] http://arxiv.org), but does not provide an explicit statement about releasing source code for the methodology or a link to a code repository. |
| Open Datasets | Yes | To validate on real probe vehicle traces, we first used GPS logs of cabs operating in the city of Porto, Portugal. The data was originally released for the ECML/PKDD data challenge 2015. |
| Dataset Splits | No | The paper states, 'We trained on the best (in terms of the number of trajectories) 24 Fridays... We tested the learnt parameters on two Fridays,' describing training and testing sets, but does not provide specific details on dataset splits (e.g., percentages, counts, or a separate validation set). |
| Hardware Specification | No | The paper does not provide any specific hardware details such as CPU/GPU models, memory, or cloud compute resources used for running the experiments. |
| Software Dependencies | No | The paper describes the algorithms and models used (e.g., 'Noisy OR CPD', 'DBN', 'EM based algorithm', 'particle filtering approach'), but does not specify any software dependencies with version numbers (e.g., Python, TensorFlow, PyTorch versions or specific library versions). |
| Experiment Setup | Yes | We chose 8 probe vehicles to circularly ply around the north-south region while another 8 along the east-west corridor. We quantified short by links < 75m in length and pick Δ = 5 min. Trajectories from 4 p.m. to 9 p.m. were considered. ... We trained on the best (in terms of the number of trajectories) 24 Fridays. |