Emergency Response Optimization using Online Hybrid Planning

Authors: Durga Harish Dayapule, Aswin Raghavan, Prasad Tadepalli, Alan Fern

IJCAI 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We evaluate the performance of three online planning algorithms based on hindsight optimization for HSA-MDPs on real-world emergency data in the city of Corvallis, USA. The approach takes into account and respects the policy constraints imposed by the emergency department. We show that our algorithms outperform a heuristic policy commonly used by dispatchers by significantly reducing the average response time as well as lowering the fraction of unanswered calls.
Researcher Affiliation Collaboration 1 School of EECS, Oregon State University, Corvallis, OR, USA 2 SRI International, 201 Washington Rd. Princeton, NJ08540
Pseudocode No The paper describes algorithms and formulations using text and equations (e.g., 'Hindsight Optimization (HOP) [Chang et al., 2000; Chong et al., 2000] is a computationally attractive heuristic for online action selection'), but does not include any explicitly labeled pseudocode or algorithm blocks.
Open Source Code No The paper includes a tinyurl link in a footnote for the 'RDDL Encoding' in Section 4, but this is specific to the domain representation, not the general open-source code for the entire methodology (e.g., the HOP implementation or experiment setup).
Open Datasets No The paper states: 'The dataset has historical emergency calls from 2007 to 2011( 1825 days, 35,222 calls), each call consists of (x, y) :location, t : time, and severity : type of the emergency.' and 'The data used for generating futures is sampled from 2400 calls between June 1 to Dec 31, 2010 unless otherwise stated.' However, it does not provide concrete access information (e.g., a link, DOI, or formal citation with authors and year for a public repository) to this 'emergency data for Corvallis, USA' dataset.
Dataset Splits No The paper mentions 'fixed test data' and 'data used for generating futures' (which can be inferred as training data for the model's future sampling) with specific call counts and dates. However, it does not explicitly describe a separate 'validation' dataset split used for hyperparameter tuning or model selection.
Hardware Specification No The paper mentions that 'Gurobi (version 7.5) solves the resulting MILPs optimally', but it does not specify any details about the hardware (e.g., CPU, GPU models, memory) on which these computations were performed.
Software Dependencies Yes We found that for small values of h and F, Gurobi (version 7.5) solves the resulting MILPs optimally for many states in under two minutes.
Experiment Setup Yes Each evaluation has three experimental parameters: (1) time limit t per decision in seconds, (2) the lookahead depth (h) = the length of sampled futures, and (3) the number of sampled futures (F) per decision. We used 5 responders for this experiment, namely a Command, Engine, Ladder and two Ambulances, lookahead h = 4 and F = 5 futures with 120 seconds being the maximum time allowed for optimization.