Challenges and Opportunities in Crowdsourced Delivery Planning and Operations
Some of the most visible and impactful societal changes of the last decade are the rapid evolution of the shared and gig economy. Companies at the forefront of these changes are Airbnb and Uber. Their business models have fundamentally changed our society. We focus on one aspect of the evolving gig economy: crowdsourced delivery. How to best deliver goods to consumers has been a logistics question since time immemorial. However, almost all traditional delivery models involved a form of company employees, whether employees of the company manufacturing the goods or whether employees of the company transporting the goods. With the growth of the gig economy, however, a new model not involving company employees has emerged: crowdsourced delivery. The Oxford dictionary defines crowdsourcing as “the practice of obtaining information or input into a task or project by enlisting the services of a large number of people, either paid or unpaid, typically via the internet.”
Crowdsourced delivery, therefore, involves enlisting individuals to deliver goods and interacting with these individuals using the internet. In crowdsourced delivery, the interaction with the individuals typically occurs through a so-called platform. A prototypical example of such a platform is the one provided by Grubhub, which links restaurants, diners, and individuals willing to deliver meals from a restaurant to a diner. The platform handles everything from facilitating the ordering of meals, to the scheduling of the delivery of the meal, to the associated payments (collecting payments for meals, distributing payments to restaurants, and distributing payments to crowdsourced drivers). Importantly, the crowdsourced drivers are not employed by the platform or by the restaurants. Crowdsourced delivery has fundamentally changed the planning and execution of the delivery of goods: the delivery capacity is no longer under (full) control of the company managing the delivery. This implies that certain aspects of goods delivery that were simple and straightforward in the traditional model are no longer so simple and straightforward. How do you plan when delivery capacity is uncertain? How do you execute when delivery capacity is uncertain? How can you ensure that you meet your service promises to your customers? Does it make sense to rely on (only) crowdsourced delivery capacity? These, and many other questions will be raised and partially answered in this presentation.
ISyE, Georgia Tech
From Learning to Optimize to Learning to Explore
We consider a discrete combinatorial space and a given objective function where the goal is not to find the maximum of the objective function but rather to discover its main modes, which can be turned into the question of sampling values with probability proportional to the objective function. By taking a power of the objective function, that formulation can smoothly transform the problem of finding the leading modes (with more or less emphasis on the really larger ones) into focussing on just the argmax of the objective. This problem comes up in drug discovery and material discovery tasks, where the objective function is only a proxy (e.g. from a simulator, or imperfect assays) for what we really care about (e.g., more expensive assays, like with mice models, or even clinical trials). Finding a diversity of good solutions is therefore important, because the single argmax solution may not in the end be appropriate. Although MCMC methods can in principle be used for that, we present an alternative approach based on deep generative models seen as policies sampling a sequence of discrete actions and that has the potential to use the power of systematic generalization in order to guess the presence of isolated modes of the objective function. This avoids the mode mixing issue which often comes up with MCMC in high-dimensional spaces where local search methods get stuck and even annealing is not enough, but instead relies on the potential of machine learning to generalize out-of-distribution, a rapidly expanding area of research in deep learning.
Computer Science, University of Montreal
Role of Optimization in Managing Amazon’s Supply Chain
Amazon runs a complex supply chain to manage the journey of each unit of inventory from the warehouses of the vendors to the hands of the customers, as the inventory passes through cross-docks, fulfillment centers, and delivery stations. At each step of this journey, optimization models play a critical role. In this talk, I will give an overview of these optimization models in a way that is biased towards my personal experience. The models operate on different scales in terms of granularity of time, geography, and product groups, which make them particularly difficult to coordinate. Thus, coordination will be a prevalent theme throughout the talk. I will conclude with a more technical discussion based on models that have been abstracted from my work at Amazon.
Cornell University and Amazon
Improving Supply Chain Resilience: Looking Back and Looking Forward
Prolonged shortages of PPE, vaccines, and semiconductor chips during the Covid-19 Pandemic exposed the vulnerabilities of global supply chains. In this plenary talk, I share my observations and discuss potential steps that government representatives, industry leaders, and INFORMS members can take to improve supply chain resilience.
Operations Management, UCLA
Challenges in the Application of Mathematical Programming Approaches to Enterprise-wide Optimization of Process Industries
Enterprise-wide optimization (EWO) is an area that lies at the interface of chemical engineering and operations research, and has become a major goal in the process industries due to the increasing pressures for remaining competitive in the global marketplace. EWO involves optimizing the operations of supply, production and distribution activities of a company to reduce costs and inventories. A major focus in EWO is the optimization of manufacturing plants as part of the overall optimization of supply chains. Major operational items include production planning, scheduling, and control.
This talk provides an overview of major modeling and computational challenges in the development of deterministic and stochastic linear/nonlinear mixed-integer optimization models for planning and scheduling for the optimization of plants and entire supply chains that are involved in EWO problems. We address the following major challenges in this area:
- multi-scale optimization
- linear vs. nonlinear models
- handling of uncertainty and disruptions, and
- multiobjective and multilevel optimization.
We illustrate these challenges in areas such as planning and scheduling of batch plants, simultaneous optimization of supply chain planning with inventory policies, optimization of business transactional processes in digital supply chains, demand side management in power intensive processes, development of infrastructure for shale gas production, and design of resilient and responsive supply chains for chemical products. These problems, which have been addressed in collaboration with industry through a consortium, have led to substantial economic savings.
Carnegie Mellon University, Rudolph R. and Florence Dean University Professor, Chemical Engineering
UC Berkeley, Department of Industrial Engineering and Operations Research
Harvard, Gordon McKay Professor of Computer Science and Director of the Center for Research in Computation and Society
Cornell University, School of Operations Research and Information Engineering
A Dynamic Queueing Road Map from Communication Systems to Resource Sharing Services
The field of operations research applies mathematics to the creation of quantitative languages designed for strategic decision making. Queueing theory was invented just over a century ago to design efficiency into communication systems. In the 21st century, it plays this same role in the design of resource sharing services. Rates for customer service demand can easily be dependent on the time of day, week, or seasonal effects. Hence dynamic rate queues are more realistic stochastic models than their traditional constant rate counterparts. Moreover, since they are not amenable to classical steady state analysis techniques, dynamic rate problems lead to greater mathematical challenges. Along with many collaborators, this talk covers a personal research journey to develop a dynamic rate queueing theory. We also show how the guideposts for our path evolved from communication systems to resource-sharing services.
Princeton University, Edwin S. Wilsey Professor of Operations Research and Financial Engineering
Policy Modeling For SARS-CoV-2 Screening, Prevention, And Vaccination
In 2020, the SARS-CoV-2 pandemic repeatedly forced decision makers to confront the tradeoff between clinical, epidemiological, and economic considerations. More often than not, a policy response was demanded long before major uncertainties could be resolved via the traditional forms of health and medical investigation. Four dilemmas will be presented, each of which was the topic of front-page media coverage, and describe Dr. Paltiel’s personal experiences developing simple policy models to address them:
- How to reopen college campuses safely?
- How to trade vaccine efficacy against speed of implementation?
- How to choose between single- and two-dose vaccines?
- How to evaluate the costs and benefits of population-wide, home-based, rapid, antigen testing?
A. David Paltiel
Yale School of Management, Professor of Public Health
Lagrangian Control at Large and Local Scales in Mixed Autonomy Traffic Flow
This talk investigates Lagrangian (mobile) control of traffic flow at local scale (vehicular level). The question of how will self-driving vehicles will change traffic flow patterns is investigated. We describe approaches based on deep reinforcement learning presented in the context of enabling mixed-autonomy mobility. The talk explores the gradual and complex integration of automated vehicles into the existing traffic system. We present the potential impact of a small fraction of automated vehicles on low-level traffic flow dynamics, using novel techniques in model-free deep reinforcement learning, in which the automated vehicles act as mobile (Lagrangian) controllers to traffic flow.
Illustrative examples will be presented in the context of a new open-source computational platform called FLOW, which integrates state of the art microsimulation tools with deep-RL libraries on AWS EC2. Interesting behavior of mixed autonomy traffic will be revealed in the context of emergent behavior of traffic: https://flow-project.github.io/
UC Berkeley, Civil Engineering/Aero Engineering
IFORS Distinguished Lecturer
2021 Wagner Prize Winner Reprise
The Daniel H. Wagner Prize is awarded for a paper and presentation that describe a real-world, successful application of operations research or advanced analytics. The prize criteria emphasizes innovative, elegant mathematical modeling and clear exposition. To learn more about the prize, visit the information page.
Boundary-Expanding OR/OM Research
OR and OM have brought about significant improvements to operations in diverse domains, including military, manufacturing and service, and the knowledge economy. Every technological advance in the modern world has been met with the pursuit of new models by the OR/OM community, often providing fundamental understanding of and significant improvements to its deployment. In this talk, the speaker will share her experience in pursuing research in the boundaries of operations and finance, wireless communications and blockchains, including the inspirations, execution, challenges and lessons learned. Pursuing such projects is not without risk, but is an effective way for a researcher to reinvent him/herself and have a fulfilling career.
Hong Kong University of Science and Technology, Industrial Engineering and Decision Analytics
Operational Data Driven Interventions to Decrease Adverse Events Associated with Opioid Overdose
In this talk, we present a systematic data driven approach to decrease adverse events associated with overdose episodes.
We take a three fold approach. First, we examine pathways that result in opioid use and devise protocols to decrease the number of new users.
Second, we predict adverse occurrence of adverse episodes among current users and adopt timely interventions that will decrease the likelihood and severity of an event. Third, we focus on the care pathways for existing users and use simple operational techniques to increase the system’s capacity as well as improve outcomes.
University of British Columbia, Operations Management